Home

DPL 8 Professional

Contents

1. Figure 9 19 Model with 2 x 25 Value Node gt Create an Influence arc from Yearly Growth to Total NPV Your model is set up so that DPL exports the 50 values in the Yearly Growth node to Excel You will now run the model gt Make sure Risk Profile Init Dec Alts and Policy Tree are selected gt Click Home Run Decision Analysis or press F10 Note unless you have changed some of the input data your results should be the same as those from model runs in earlier sections 275 Chapter 9 Multidimensional Value Node Syncopation Software This tutorial has shown you how to set up and link a two dimensional value node in DPL In this example the values in the node were constants representing growth rates However each element of the node could have been set up as any expression using DPL variables or functions Chapter 1 of this manual includes more information on how to set up Get Pay expressions in DPL when an array or series is used to define the Get Pay as well as an example using DPL s Time Series Percentiles feature in conjunction with a one dimensional value node 2 6 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics 10 TIME SERIES PERCENTILES AND MULTIPLE METRICS Time Series Percentiles are useful for understanding how the range of uncertainty in performance metrics such as gross sales or net profits evolves over time This chapter discusses how to run Time Series
2. LUUL Q VY General Conditioning Print Page Setup Copy FullScreen Wheel Reorder Q Q Q ee DO Links Node Definition p Figure 3 35 Node Definition Dialog Data Tab for p Click OK Double click the Cost per Unit node to edit it The Node Definition dialog appears with the Data tab selected In place of 0 6 for the probability of Nominal given In house type p Press the down arrow twice to select the probability for High Delete the 0 4 that is there for the Probability of High Press Enter Your node data for Cost per Unit should now look like Figure 3 36 118 Syncopation Software Node Definition Cost per Unit Chapter 3 Decision Analysis Results General Conditioning Print Page Setup Copy FullScreen Wheel Reorde QQA Daa Distribution type discrete a Links Probability vf Value vif A Cost Nominal In house per Unit Product Strategy Cost License per Unit 5 0 y lt gt Cancel Hep Figure 3 36 Node Definition Data Tab for Cost per Unit Note in the changes you just made you left the probability of the last outcome of the discrete chance node blank You can always do this DPL will assign one minus the sum of the rest of the outcome probabilities to the last outcome In this instance you must do this Rather than entering a number for the probability of Nominal you ha
3. a x 42 1 STS foom Figure 12 9 Scenario Risk Profile Chart You will see a Risk Profile Dataset and a Risk Profile Chart item both called Scenario Expected Value Note Scenario Risk Profile charts and datasets will overwrite Expected Value charts and datasets Since you renamed the Expected Value dataset to Full Profile it did not get overwritten As you would expect the risk profile and expected value for this scenario are very different from the full profile because you have removed the possibility that the product is not approved You can compare the Scenario Risk Profile chart with the Full Profile chart you created and see the differences Also note that the Scenario Risk Profile has a title indicating the states of the events that are not varying i e Develop and FDA Approval are both set to Yes You will generate one more Scenario Risk Profile using the dialog gt Open the Policy Tree again gt This time right click on the chance node for Key competitor outcome and select Scenario Risk Profile from the context menu to generate the dialog gt Use the combo box to select the High setting for Market Size See Figure 12 10 350 Syncopation Software Chapter 12 Advanced Decision Analysis Results Scenaro Risk Profile Drug Development Bars Develop_ FDA_Approval Key_competitor_outcome All Launch CAINy Market Share FAID Mlake Size i a Eee oe Da o Pricing
4. Prefix node name with sheet name Create DPL program export nodes Figure A 6 Range Names Dialog for Creating Initialization Linked Valude from Excel Using this method DPL creates value nodes for the selected ranges If you have Create arrays based on range size DPL will create one or two dimensional arrays to match the range size 2 To establish initialization links for an existing node do the following In the Influence Diagram pane edit the node that you wish to link gt Switch to the Links tab Turn on initialization links by clicking Microsoft Excel in the Initialization Links section 446 Syncopation Software Appendix A Overview of Spreadsheet Linking gt If needed browse for the spreadsheet workbook you wish to use for the link Note that in the dialog shown in Figure A 7 the initialization link is established with a node that already has calculation links The Same as calc links option is checked so you are linking this node to the same workbook However if the node did not already have calculation links or if you want to link it to a different workbook for initialization you would uncheck the Same as Calc links option and use the Browse button to find the new workbook Node Definition Marketing phase 2 General Calculation links These links are recalculated throughout the run They may be either inputs used in the linked spreadsheet s calculations e g unit sales or c
5. Zoom in buttons to increase decrease zoom Zoom Slider Control level or manually click and drag the Zoom slider control LJ to the desired zoom level Table 1 1 Zoom Controls on the Status Bar You can also zoom these windows by Ctrl Shift selecting To zoom ona specific region click the left mouse button on the top left corner of the region you d like to view hold Ctri Shift and then drag the mouse down and to the right A rectangle indicates what will be displayed when you release the button Release the mouse button when the rectangle surrounds the portion of the diagram you wish to view 11 Chapter 1 Introduction Syncopation Software 1 4 6 Introduction to the DPL Command Ribbon Above the Workspace and Model Windows you will find a Microsoft Office style command ribbon consisting of multiple tabs Each tab is split up into contextual groups that contain one or more commands You can view a short description of a particular command on the ribbon by holding your cursor over the item until the tooltip note appears File Tab The File tab contains standard commands for creating a new Workspace opening and saving Workspaces and printing You can also access DPL and Workspace level options via File Options These options will be discussed in more detail later in the chapter Home Tab The Home tab contains a Workspace group within which you can add delete rename copy cut paste select and find replace
6. Mkt_1__ Mkt_1_an Mkt_1_an Mkt_1 and_2__30 BCM Include_Market_2 Yes No Pipeline Capacity BCM_10 BCM_20 BCM 30 CapEx Low Nominal High Actual_Price Low Nominal High Actual_Reserves Low Nominal High Market_2 Political Sta Stable Unstable Forecast_Price Low Nominal High Market_2 Political Sta Stable Moderate Unstable Operating Costs Low Nominal High 31 Appraisal_Results Low Nominal High M 4 gt gt Endpoints2 J i 4 ul Eo 100 wan auf wn Figure 11 24 Endpoint Database in Excel with State Numbers 333 Chapter 11 Using the Endpoint Database Syncopation Software Note that at the top of this file there are several comment lines explaining the file format This format is required in order for DPL to re import the endpoint data As you can see in Figure 11 24 the file other than the comment lines is very similar to the previous file you exported however instead of state names the data other than the last two columns which are the probabilities and the objective function values indicates the settings of each state using state numbers The state numbers are indexed beginning with zero for each event If you should need to import endpoint data that was created by an application other than DPL such as a statistical package or just within Excel you will need to make sure that the data is in the correct format The easiest way to do this is to export a database from DPL using state numbers as you
7. Show cells with formulas Figure A 4 Cell Names Dialog when Linking an Existing Node 443 Appendix A Overview of Spreadsheet Linking Syncopation Software Node Definition FDA Approval General Calculation links These links are recalculated throughout the run They may be either inputs used in the linked spreadsheet s calculations e g unit sales or calculated results from the Data spreadsheet e g NPV None local DPL Program Microsoft Excel Workbook Asymmetric Trees xls Browse Sheet Cell Assumptions FDA_approval Go to Cell Cell Names y This is an export node Its value will be sent to Excel each time it changes during the run Paste Link Initialization links These links are for connecting to a spreadsheet database or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at the start of each run None local or DPL program C Microsoft Excel Database Figure A 5 Cell Names Dialog when Linking an Existing Node With this method DPL will also adjust which cell names appear in the dialog depending on whether the node you are linking has data in it If it has data you will see that Show cells with formulas is not checked because DPL assumes it will be an export node Conversely if the node has no data by default DPL checks Show cells with formulas because DPL assumes the node will be an im
8. The decision tree in Figure 4 1 has over 43 million paths It is likely that not all of these paths make sense For example the company may not want to add a number of new features to all of their products and cut prices on all of them Assume the company is considering a number of alternative strategies for its future product development You will create a strategy table to model these development strategies 126 Syncopation Software Chapter 4 Strategy Tables gt Click the Zoom Out button on the status bar to make the Influence Diagram smaller and create space above the decision nodes gt Click Model Node Add Strategy Table on the ribbon see Table 1 2 gt Click above the decisions in the Influence Diagram as shown in Figure 4 2 Figure 4 2 Cursor for Strategy Table You will be asked whether you want to add decisions to the strategy table as shown in Figure 4 3 If you answer yes DPL will create the strategy table with the decisions you select If you answer no DPL will create the strategy table with two new default decisions Figure 4 3 Add Existing Decisions to Strategy Table Question 127 Chapter 4 Strategy Tables Syncopation Software Click Yes The Select Decisions dialog appears as shown in Figure 4 4 Select 2 or more Decisions Feature Set 4 Feature Set E Feature Set C Marketing Spend Price Change 4 Price Change E Price Change C Figure 4 4 Select Decisions Dialog
9. lt 100 gt 49 100 lt 0 gt lt 0 gt 52 0 does not occur lt 0 gt 16 does not occur lt 0 gt 0 does not occur lt 0 gt 0 51 17 lt 32 gt Drilling Costs Oil Price Pipeline Costs Low lt 20 gt Low lt 15 gt Low lt 30 gt 10 30 30 lt 70 gt i lt 49 gt i 34 40 High lt 10 gt lt 36 gt 60 lt 10 gt 5 30 10 does not occur lt 0 gt does not occur lt 0 gt does not occur lt 0 gt 51 0 0 Seismic Structure Exp Seismic Test No lt 0 gt No lt 0 gt 41 lt 51 gt 35 lt 49 gt 24 lt 0 gt 0 does not occur Figure 12 14 Policy Summary with Comparison Bars for Drill Yes In comparison mode DPL displays graphically the policy dependent probabilities for the full set of optimal endpoints with a blue bar below each branch It also displays graphically the policy dependent probabilities for the filtered endpoints with a red bar above each branch DPL displays the numeric values for the filtered policy dependent probabilities in angle braces e g lt 32 gt above the end of each branch As before the policy dependent probabilities for the unfiltered set of optimal endpoints are displayed below the branch Chance outcomes or decision alternatives that have a red bar that is longer than their blue bar are more likely in the filtered scenarios than they are in the full set of optim
10. 194 Syncopation Software Chapter 6 Asymmetric Trees You will now look at how to set default states gt Open the Pricing chance node in the Influence Diagram Click on the General tab The Nominal outcome is designated as the default state for Pricing Select another state and press the Default button The default changes gt Click Cancel to discard the change Each chance node and decision node should have one of its outcomes assigned as the default in the General tab of Node definition In particular if you are building an asymmetric model such as the example used here you should ensure that each node has a default state set You will now run an analysis gt Click Home Run Decision Analysis or press F10 The results are the same as before although the tree is now much smaller and the analysis runs more quickly To see how many paths there are in the model you can look in the session log Look for the Number of paths entry near the bottom The end of the session log will look something like this 11 00 00 Compiling Asymmetric Number of paths 29 11 00 01 Complete 11 00 01 Analyzing Full tree enumeration 11 00 02 Complete The tree now has only 29 paths compared with 108 when you started Since the tree is quite small either way this may not seem to make much difference but in much larger models incorporating asymmetries in the tree will result in improved performance and clearer c
11. Daia Ten Year Hide Scalar value Cash Flow i Links 1 dimensional aray N values or series C 2 dimensional array M x N values DPL variable name for expressions Ten_Year_Cash_Flow Comments Print All Comments Copy to Clipboard Carat Hb Figure 9 5 Node Definition Dialog for 1 Dimensional Array gt Switch to the Data tab of the Node Definition dialog Note that DPL has created a series by setting the Series interval entry to be true And it has created a single interval which goes from 0 to 9 So DPL has created a single interval series with ten elements See Figure 9 6 This matches what DPL found in Excel when it examined the named ranges e g that Ten_year_CF is a ten cell named range with formulas DPL will create a series import node if it finds formulas in the named range It will create an array export node if it finds numbers in the named range You will learn more about series interval entry in the next section 260 Syncopation Software Chapter 9 Multidimensional Value Nodes General Intervals od Series interval entry _ Create Mx 1 column vector v Interval From To Expression 0 0 9 Figure 9 6 Data Tab for Single Interval Series gt Switch to the Links tab of the Node Definition dialog Note that DPL has linked the node to Ten_year_CF See Figure 9 7 General Calculation links These links are recalculated throughout the run They may be either inpu
12. Select all the decisions in the list gt Click OK The Node Definition dialog appears as shown in Figure 4 5 Node Definition Strategy Table Data Strategy Hide Table DPL variable name for expressions Strategy_Table Figure 4 5 Node Definition Dialog General Tab for Strategy Table 128 Syncopation Software Chapter 4 Strategy Tables gt Change the name of the strategy table to Strategy Select the Data tab As shown in Figure 4 6 the Data tab displays the Strategy Table Window a Node Definition Strategy General Commands Print Page Setup Copy Format Full Screen oa Feature Feature Feature Marketing Price Price Price Set A Set B Set C Spend ee oe ir E Strategy1 Add features Add features Add features Increase Increase Increase Increase F Strategy2 Keep same Keep same Keep same Keep Same Keep same Keep same Keep same Minimal Minimal Minimal Decrease Decrease Decrease Decrease Figure 4 6 Data Tab for Strategy Table In the leftmost column of the Strategy Table Window DPL displays the strategies defined in the table By default DPL creates two strategies named Strategy1 and Strategy2 DPL also uses colored icons to represent strategies The first strategy in the strategy table is always represented by a red square icon In each of the following columns DPL displays the decisions included in the strategy t
13. 369 Chapter 13 Risk Tolerance Syncopation Software Sales In house 29 3 Product Strategy 29 3 Sales License 28 7 Figure 13 5 Policy Tree with Certain Equivalent results In house is still the preferred alternative but the gap has narrowed The results in Figure 13 5 are Certain Equivalents which are being shown because risk tolerance is now incorporated in your model Pursuing the In house alternative might result in an outcome better or worse than 29 3 but a decision maker with a risk tolerance of 50 would be indifferent between the investment and a perfectly safe 29 3 If as is usually the case the decision maker is risk averse the Certain Equivalent will be less than the Expected Value If you would like to see the Expected Value results Figure 13 2 again click View Results CE to turn off the Certain Equivalents view DPL changes the Policy Tree to show the EV results In a model with a risk tolerance DPL always calculates both EV Expected Value and CE Certain Equivalent results With a risk tolerance coefficient of 50 the optimal decision is unchanged How much lower would it have to go i e how much more risk averse would you have to be for the License alternative to have a higher certain equivalent than the In house alternative gt Click Home Sensitivity Rainbow Diagram gt Click Select Select Risk_Tolerance from the Value for sensitivity drop down list Clic
14. Analysis _ Init Dec Alts O vor C Value Correlations Run Figure 7 2 Decision Analysis Options dialog The Policy Tree tells you that you should drill and that the expected value of this opportunity is 26 5 million Figure 7 3 Drilling Costs Drill Yes 26 9 26 5 No 0 0 Figure 7 3 Policy Tree 217 Chapter 7 Conditioning and Learning in Decision Models Syncopation Software gt Navigate to the Risk Profile Chart double click on the item for it in the Workspace Manager or click on the item for it in the document navigator gt 5 Q p a ov 2 ad L 5 E 5 O 90 60 30 0 30 60 90 120 150 180 210 240 270 Figure 7 4 Risk Profile While this opportunity has positive value it is risky There is a about a 51 chance of losing money see Figure 7 4 For more on reading Risk Profiles see Section 3 2 218 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models In reality the wildcatter has more choices than just whether or not to drill S he can conduct one of several tests intended to tell him r about the likelihood of finding oil and then decide whether or not to drill You will add a decision and a learning uncertainty for one such test gt Add a decision node to the Influence Diagram Place it above the Drill decision Call the decision node Test Name its alternatives None and Core sample Add a
15. Chapter 1 Introduction Syncopation Software Fang mel Excel Calculation Linked Add Linked Node i i Excel Initialization Linked Add linked node of default Brei R DPL Program Calculation Linked type as indicated by Icon see icons below Use drop Links m e Database Initialization Linked down split to select link type The dialog box launcher will open the other than default The Model Links dialog where you can view default changes to the last model links specify DPL program links added linked node type convert a spreadsheet to a DPL Program and change links settings in Excel Calc Link mpr 2 Excel Init Link DPL Program Calc Link i X ta Database Init Link Add New Arc Add new influence arc between nodes From Formulas ba Influence Diagram that are e Se a B based on the formulas of the Formulas a definitions of each node E Bend Add new influence arcs to amp E Influence Arc Straighten All Bend Straighten Bend or straighten the selected influence arc To straighten all arcs at once drop down Straighten split button and select Straighten All Table 1 3 Model Links and Model Influence Arc Ribbon Groups 14 Syncopation Software Chapter 1 Introduction Default Tree Build or re build the default tree The existing tree will be deleted Add Instance Add a new node instance decision or chance to decision tre
16. Current value Figure 5 13 Completed Value Tornado Setup Dialog Click OK DPL displays a Value Tornado diagram Figure 5 14 for the eight value nodes tested The bars are sorted so the widest one corresponding to the highest impact value is at the top This diagram will help decide which value drivers need to be modeled probabilistically as chance nodes See Section 2 5 for more information on Value Tornadoes 155 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Value Tornado Gross Margin Trendline Revenue Growth Rate Share of at Risk Revenue Lost RMC Lawsuit Settlement Bad Debt Allowance Share of Revenue at Risk Working Capital Other Cost Growth Rate 500 600 700 800 900 1000 1100 1200 Expected Value Low Policy Change Policy same as Nominal High Policy Change Figure 5 14 Value Tornado Diagram 156 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models 5 5 USING THE WORKSPACE MANAGER DPL allows you to store any number of Models in a Workspace Before continuing you ll make a copy of your model as it now stands so that you ll have it if you need to update the tornado diagram later Most actions in the Workspace Manager Window are accomplished by right clicking an item and choosing a command from the context menu First you will make a copy of the current model gt In the W
17. High High High No 1 lt PV Market Figure 6 28 Decision Tree with Launch Decision node moved You now need to make the launch decision asymmetric Double click on the Launch decision in the Decision Tree The Tree Instance tab of the Node Definition dialog appears Select Asymmetric for the Alternative Grouping Click OK Drag the Market Share node and attach it to the Yes branch of the Launch decision Your decision tree should look like Figure 6 29 GU Market Market cite Share Size Pricing puicome Launch FDA Low Low Low Approval Fi A aT a i EOD haden Develop JEn ff PN Market Yes Nominal Nominal Nor nal a g Kominal E PV Market es A ay Launch og High _High High Pv _Devwelopmeni No No High PW Market No a Figure 6 29 Decision Tree Attached 208 Syncopation Software Chapter 6 Asymmetric Trees Finally you need to make sure that the launch costs are only paid if you decide to launch the whole point of the launch decision is to avoid those costs if the project looks unattractive gt Select the Yes branch of FDA Approval gt Delete the Get Pay expression in the Model Get Pay drop down and press Enter gt Drag the PV_Launch value node to the Yes branch of Launch in the Decision Tree pane and release the mouse button Ke Market Market compbttot Share Size Pricing outcome _ Launch Low Low Develo Failure Low
18. are checked Run a Full Tree Enumeration from Endpoints You can see from the Policy Tree that the expected value of the decision has declined to only about 4 5 You can compare the what if Policy Summary with the one you saved Original and see that the optimal initial decision alternative has changed to the Low level of investment and that the policy dependent probabilities of the downstream strategy have changed significantly as well For example the No branch of the Proceed decision is chosen much more often gt Switch to Model window and click Model Branch Uncontrol All to return the model to its earlier state Delete Original Policy Tree by right clicking its item in the Workspace Manager and selecting Delete gt Delete Original Policy Summary 326 Syncopation Software Chapter 11 Using the Endpoint Database 11 3 RECONNECTING ENDPOINTS TO A MODEL In some situations you may find it useful to rename and save an Endpoint Database just as you can rename and save other model outputs in DPL When you save an Endpoint Database you can later reconnect your model to the saved database and play its endpoints You might need to do this for example if you make changes to your value model e g in your linked Excel spreadsheet but then wish to revert to the original version to compare results You will do a simple example of this now Double click on the Endpoint Database icon
19. cccceeeees 245 NAMING MOEIS cccceeeeeseeeeeeeees 32 Node Definition dialog data input tree ccccseceeseeeeeeees 28 node types Mang meira 59 number of intervals ccscseceeeee 290 480 Syncopation Software Objective GialOG cccsccscsseeeeeeees 242 Objective FUNCTION ccceeeeeeeeeees 235 Option Value Chart scceceeeeees 361 default States i ccecescsesereeees 362 options directive cceceeeeeeeeeeeeees 423 pay keyword seseseseseserenererererenene 417 perform keyword ccsceeeeeeeeeeeeees 416 Perform Reference sccseseesereeees 228 perform subtree ccceceeeeeeeeeeees 227 Perform TarGet cccccscsseseeseeeeeees 228 policy dependent probability 106 Policy Output Options s sesesesesesenen 76 Policy Summary asasen 91 106 352 ComMmpariS n oaran 355 Policy Tree assssssssssssrssssssens 78 341 expand tO leVel ccecseceesereeees 190 FIENO arora 344 Format Display ccccscsseeeeees 84 INCFOCUCTION ecceceeeeeeeceeeeeeeeees 52 DFECECESSOM cscsseseeseueeeeeuseeeeeans 100 quit keyword ccceceeeeeeeeeeeeeeeeees 416 Rainbow Diagrams 6 113 248 policy changeS sseseseseseseserens 381 real OPtiON ena a 203 reserved word sesesesesesesenenenenenns 394 results View EV CE sssssessssssesensrsrrens 370 Risk Profile ENa Sna antecene ANoataues 85 cumulative ssssssssssssnsrsrsrsrenenns 86 datasets ienie E 85
20. character indicates that the Medium state is the default state The tilde is optional however it is recommended if you will be using Base Case Tornado diagrams or Asymmetric Tree Dependency Checking If no state is marked with the tilde the first state is the default If you wish to associate value expressions with the states of a decision add them to the definition following the state names decision Research Funding Low Medium High 10000 25000 35000 15 4 5 Chance Events The following statement defines a chance event in DPL chance Research Time short Medium Long Og pu Ove In the statement above only two probabilities are specified If no probability is given for the last state DPL will calculate it as one minus the sum of the other probabilities Like decisions chance events can have default states indicated by the tilde character 400 Syncopation Software Chapter 15 DPL Programs and Code If you wish to associate value expressions with the states of a chance event enter them after the probability distributions A chance event with both probabilities and values assigned to its states is sometimes called a random variable chance Research Time Short Medium Long 0 3 0 5 1 5 10 in months Chance events may be drawn from named distributions To do so do not enter probabilities and value expressions Instead use the distribution name and enter value expressions for the param
21. l Figure 13 1 License vs In house Model Run the model to review the current results gt Make sure Risk Profile In Dec Alts and Policy Tree are checked in the Home Run group Select Home Run Decision Analysis or press F10 366 Syncopation Software Chapter 13 Risk Tolerance In house 32 9 Product Strategy 32 9 License Figure 13 2 Policy Tree The Policy Tree Figure 13 2 shows that In house is the optimal alternative The expected value of In house is 2 9 million more than the expected value of License Navigate to the Risk Profile Chart called Initial Decision Alternatives gt Uncheck EV CE Lines and EV CE Values in Chart Format Display 367 Chapter 13 Risk Tolerance Syncopation Software Cumulative Probability 100 95 90 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 10 5 In house License Figure 13 3 Risk Profiles The Risk Profiles of the two alternatives show a more ambiguous result Figure 13 3 In house is better in terms of expected value but it is also more risky Would a decision maker s attitude toward risk ever lead him r to choose License gt gt LUULU Return to the active model Add a Value node to the model by clicking Model Node Add add value should be the current default see Table 1 2 The new node will contain data for your risk tol
22. 100 License 0 does not occur 0 lt Figure 14 9 Session Log with Policy Information from Two Way Rainbow Diagram Return the Session Log to its default size 383 Chapter 14 Sensitivity Analyses Syncopation Software 14 2 EVENT TORNADOES Event tornadoes provide information about how much each chance node in a model contributes to the overall uncertainty in the outcomes of the objective function of the model There are two types of Event Tornadoes in DPL Deterministic and Probabilistic DPL creates a Deterministic Event Tornado by running the model once to establish the minimum amount of uncertainty in the model called the Minimum Model Range DPL calculates the Minimum Model Range by replacing all the chance events in the model with their expected values Then DPL runs the model one additional time for each chance event in the model call this event the sensitivity event In each of these subsequent runs DPL replaces all of the chance events in the model except the sensitivity event with their expected values It leaves the sensitivity event unchanged In effect in each sensitivity event model run the only uncertainty left in the model is the uncertainty associated with the sensitivity event unless Always Gamble has been set for chance events see below DPL follows an analogous procedure to create a Probabilistic Event Tornado DPL runs the model once making no changes to the model to establish the original
23. 160 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models Fi Risk Profile E Policy Tree Levels none 7 Monte Carlo Samples Objective Function _ Policy Summary T Endpoints Initial 100 Decision Analysis LJ Init Dec Alts O vorc _ Value Correlations Restart 100 Run Figure 5 19 Home Run options for Monte Carlo Simulation Note that the default run type indicated in the Decision Analysis split button icon changed to Monte Carlo simulation when you introduced a continuous chance node to your model Select Home Run Decision Analysis or press F10 DPL checks your model for completeness whenever you run any kind of analysis DPL will give you an error message if you ve mistyped a formula or forgotten a value DPL runs the Monte Carlo simulation with 100 samples which is a very small number but useful for illustration When the simulation run is complete DPL displays a Risk Profile Chart Figure 5 20 The sample is random so your results will be slightly different but the general shape will be similar In this simulation each sample has a weight probability of 1 100 or 1 In the graph you can see how the histogram bars represent single samples probability 1 at the high and low extremes 161 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software gt Ta 8 Q O i A Figure 5 20 Risk Profile Histogram with 1
24. 3 has subscripts of 1 and 3 Array subscripts can also be formulas Arrowheads In the Influence Diagram the color of an arc s arrowhead indicates the type of conditioning the arc implies Black arrowheads indicate timing only blue arrowheads indicate values only green arrowheads indicate probabilities only light orange arrowheads indicate both values and probabilities Asymmetric Node A decision tree node with branches that lead to different nodes 459 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software Attribute A measure of value Profit Health Effects Environmental Effects tracked for each path of a decision tree during an evaluation Attributes are usually combined at each endpoint by an objective function generating a single measure of value that is used during tree rollback Probability distributions for each attribute and the objective function may be graphed separately B Base Case Tornado Diagram Used with probabilistic models a diagram generated by evaluating the outcome of a model when each chance event is individually set to a high and a low state while all other chance events are set to a nominal state The Base Case that the results are compared to is the result when all chance events are set to a nominal state Base Result The expected value and certain equivalent if risk tolerance is used of the objective function for the initial run of the model in a Value Tornado before any
25. 4 Ill rf Ready 23 Bam 90 g Figure 9 2 Assumptions Sheet of Multidimensional Value xls 256 Syncopation Software Chapter 9 Multidimensional Value Nodes The Assumptions sheet contains various market price and cost growth assumptions numerous other parameters and several switch cells for decisions and uncertainties Some cells are linked to DPL export nodes and the color coding indicates whether they are linked to value decision or chance nodes You can also see the two cells linked to DPL import nodes Upfront_NPV and Total_NPV You will run the model to generate results for comparison later on Switch to DPL Inthe Home Run group Fast Sequence evaluation should be the default indicated in the Decision Analysis icon and Risk Profile Init Dec Alts and Policy Tree should be selected Click the Home Run Decision Analysis icon or press F10 The model may take a couple of minutes to run Examine the Risk Profile and Policy Tree results You will now examine a named range in Excel that contains 10 cells It will be used for the value node you are going to create in DPL corresponding to the annual cash flow Switch back to Excel gt In the Assumptions sheet in Excel select the range E42 N42 This is the first 10 years of annual cash flows in the model It is labeled Total Annual Cash Flow Time Series in the Results section Note in the Name combo box
26. Click OK In this model you ll use an additive convention so all costs will be entered as negative numbers In your Influence Diagram DPL has drawn an arc arrow from Test to Test Costs to reflect the conditioning you specified in the Node Definition dialog The arrowhead is blue which means that the Test Costs have different value expressions for each state of the conditioning event Test The color codes for arcs are summarized at the end of this chapter Double click on the arrowhead DPL displays the Value Influence Type dialog Figure 7 8 221 Chapter 7 Conditioning and Learning in Decision Models Syncopation Software Value Influence Type EJ Values Different values for each conditioning event state Make this the default value influence type Cancel Figure 7 8 Value Influence Type dialog Unchecking the Different values checkbox would negate the conditioning you just added In DPL you have the choice of specifying conditioning in the Node Definition dialog when you re entering data or on the Influence Diagram itself by changing arc types If you are developing a model in which you know that influence arcs into nodes will mainly be of one type you can check the Make this the default checkbox Future arcs will be of the same type You may have a different default arc type for each node type of the successor in DPL decision chance value and controlled gt Click Ca
27. Event Data Proceed Yes No Initial_ Upstream_Investi Low High Pipeline Strategy Mkt_1 10Mkt_1_ 20Mkt_1_ and _Mkt_1 and_Mkt_1 and 2 30 BCM Include_Market_2 Yes No Pipeline_Capacity BCM_10 BCM_20 BCM_30 CapEx Low Nominal High 2 3 4 6 7 8 Actual_Price Low Nominal High Actual_Reserves Low Nominal High Market_2_Political_Stab Stable Unstable Forecast_Price Low Nominal High Market_2_Political_Stab Stable Moderate Unstable Operating Costs Low Nominal High Appraisal Results Low Nominal High Endpoint Data Proceed Initial_ Upstr Pipeline _Str lnclude_Mar Pipeline_Ca CapEx Actual_PriceA Yes Low Mkt_i1__10No BCM_10 Low Low Li Yes Low Mkt_1 _10No BCM_10 Low Low Li 4 4 gt gt Endpoints1 J i 4 T 20 Ready EB GN 110 g Yv Figure 11 23 Endpoint Database in Excel with State Names Next you will export the endpoint data again using state numbers instead of state names Switch back to DPL if you were viewing the data in Excel Activate the Endpoint Database again if needed 332 Syncopation Software Chapter 11 Using the Endpoint Database gt Export the Endpoint Database and give the file the name Endpoints2 csv this time Click Save gt Click No to the DPL prompt in order to save state numbers gt Examine the file in Excel It should look like Figure 11 24 Endpoints2 Microsoft Excel Home Insert Formulas Data Review View General g Insety E A
28. FDA Approval If it does succeed its revenues will be influenced by three factors Price Market Share and Market Size These three factors are all modeled as uncertainties Price is assumed to be set by a regulatory process so it is not under management s control in this example First run the model and look at the results gt Check Policy Tree in the Home Run group gt Uncheck all the other outputs Figure 6 2 O Risk Profile Policy Tree Levels 6 all none T Policy Summary T Endpoints Decision 5 Analysis Init Dec Alts O vorc C Value Correlations Run Figure 6 2 Home Run Group for Symetric Model 184 Chapter 6 Asymmetric Trees Syncopation Software gt Select Home Run Decision Analysis or press F10 DPL will launch Excel and open AsymmetricTrees xls From the Policy Tree Figure 6 3 you can see that the optimal alternative is to continue to develop the drug If you do so the expected NPV is approximately 75 million whereas if you abandon it the value is Zero gt Expand the Policy Tree by Shift double clicking on both instances of the FDA Approval chance node FDA Approval Yes 74 8 Develop 74 8 FDA Approval No 0 0 Yes 50 No 50 Yes 50 No 50 Market Share 323 7 Market Share 174 0 Market Share 0 0 Market Share 0 0 Figure 6 3 Policy Tree for the Symmetric Model If the decision is made to abandon th
29. Lew Liss Le Stable Morwul Komea High 01 Tau g b hiii He PCM Low Lys Le Seah minsi tagh i any ETT IT Wee Wa He BM Low Low Lew Seek iprit H aS ti a l E L Mit Hs BLM Lra E Lee Stable HomrunH Hg High sehn Tau H 1ra hiii Hic GCE Lys Low Le tabis kig T A afeh 161 t lL Wiin He BCH Li Le Low Stable Hizi rN Taraa IN T th L AM_1__ Ne BLM Low Le Lew Shaki Hih Lew Hgh eG 14 0 Tit Po MEF Lg Lin Low tabis bige Pis Toh oo 11 H Vet Lire bi He BCM ID Low Low Le jeable Hr H mae 1 16g 2S Wes Low Miti Ne BCM MS L s L s Le Snakes Hig Fiore Hy 1 1 Witt 7 __ He EAL be Lys b Tabia 7 Bagh 1 mM M Wet Lends kbit l H ECH I L s L s L s prn bigh b 27 Ves Liver Miri He ECH I Low Low Low Saabs Hig Hah Hg 1 E an Ye Lieve Wii TI Me H loa Lor bite VAmertabte ba a 1 5 H Y Lir kiia l hs ECM b L s L s Ls Livable A Tar w H Wits Liw Miir He ECH i L s L w Le Libie i H Eir t p nany Mode H j4 Undpoerd Owtabeee M i eved desipcorte W ak Figure 11 18 Workspace Manager with Saved Endpoint Database 327 Chapter 11 Using the Endpoint Database Syncopation Software At this point the renamed Endpoint Database Saved Endpoints is not connected to your model it is essentially a backup copy of your Endpoint Database Note because it is not connected to your model the Endpoint Database displayed in the report Saved Endpoints does not exist anywhere except in the report If you delete an Endpoint Database that is
30. Nominal 814 6 Nominal 838 9 30 40 40 1135 8 High 1185 8 30 1482 7 Low 653 9 Pricing 30 950 9 High 1191 9 Nominal 1224 3 30 40 1521 2 High 1686 8 30 1983 7 lt gt Figure 12 7 Policy Tree Zoomed to Expanded Subtree You can see from Figure 12 7 that there is no Low branch for Market Share because of the filter The rolled back expected value at that node is negative Similarly there is no Low branch for Market Size at the end of the Nominal branch for Market Share the rolled back expected value is negative at that point as well Lastly there is no Low branch for Pricing at the end of Market Share equal to Nominal and Market Size equal to Nominal the rolled back expected value in this case the value at the endpoint is negative The branch filtering feature can be used to give you information on which scenarios lead to negative expected values Note that if a branch does not appear because its rolled back expected value is outside the range of the filter this means that the expected value over all the scenarios further down the subtree is outside the range of the filter it does NOT mean that EVERY scenario further down the subtree is outside the range of the filter You can also filter branches based on probabilities as shown in the dialog in Figure 12 3 If you do this branches with probabilities outside the range you specify will be hidden from display The filtering is handled similarly to the roll
31. PV_Market Yes Nominal Nominal Nominal lt 1 Disappointment PV_Launch PV_Market Yes No High High High PV_Development Breakthrough PV_Market No Figure 6 30 Completed Decision Tree Your decision tree should now look like Figure 6 30 You are now ready to run an analysis Select Home Run Decision Analysis or press F10 Market Share Launch Yes 394 9 Failure 394 9 122 9 30 No 174 0 ad j f Market Share Key competitor outcome Launch Yes 323 7 FDA Approval Yes 216 9 Orsappointment 323 7 122 9 Develop Yes 21 4 x 50 Q 40 _ No 174 0 lt 21 4 174 09 Market Share Launch Yes 103 6 Breakthrough 103 6 30 No 174 0 174 0 20 lt q No 0 0 Figure 6 31 Policy Tree with Launch decision 209 Chapter 6 Asymmetric Trees Syncopation Software In the Policy Tree Figure 6 31 the expected value has changed significantly Previously you were assuming a Key competitor outcome of Disappointment but now the outcome can be either better or worse than that The bad outcome a competitor breakthrough has more of an effect than the good outcome competitor Failure You can also see that the optimal decision alternative for Launch is always Yes Although a Key competitor outcome of Breakthrough gives you a negative expected value of 104 million this is still better than quitting and losing 174 million You w
32. Sort In the Sort by drop down list select Objective Function gt Click the Descending radio button See Figure 11 7 Sort by Objective Function Descending Then by none Ascending Descending Then by none Ascending Descending Figure 11 7 Sorting the Endpoint Endpoint Database gt Click OK Note to sort by a single column it is quicker to select a cell in the column Pan you d like to sort by it will turn magenta and press the Descending cat Al or Ascending 2h button in Grid Sort amp Filter The Sort dialog is useful when you d like to sort by more than one column Scroll to the right of the report You can see that the maximum value of the objective function is 338 36 this occurs when all outcomes are favorable actual prices are High operating costs are Low and so forth 314 Syncopation Software Chapter 11 Using the Endpoint Database Select Grid Sort amp Filter Sort again and sort by none to return the Endpoint Database to its default display Endpoints in the Endpoint Database are sorted by default in the order in which DPL recorded them You can also click in the leftmost column of the report to return the sort order to the default recorded order 11 2 PLAYING ENDPOINTS If you are running a DPL model that is relatively large and is linked to an Excel spreadsheet that is also relatively large you can save time by recording and playing
33. Table 7 3 Arc Types Available Between Node Types 2353 Chapter 7 Conditioning and Learning in Decision Models Syncopation Software 234 Syncopation Software Chapter 8 Objective Function and Multiple Attributes S OBJECTIVE FUNCTIONS AND MULTIPLE ATTRIBUTES Many decisions can be effectively modeled considering only a single measure of value such as the net present value NPV of cash flows discounted at weighted average cost of capital WACC However some decision problems including many of those in the public sector may require you to quantitatively model several measures of value or you may find that including more than one measure is more enlightening These value measures are referred to as attributes Attributes can be useful even when the focus of the analysis is financial You may find it useful to look at both free cash flow and earnings for example When you wish to track financial measures or other measures over time DPL s Time Series Percentiles analysis feature lets you use one attribute for each time period to do this see Chapter 1 of this manual In large DPL models where multiple attributes are required DPL s multidimensional value node feature may also be useful see Chapter 9 of this manual The tutorial in this chapter addresses a decision in which the attributes are not all financial Chapters 9 and 1 use examples of slightly more complex multiple attribute models in which the focus is financial DPL giv
34. and the decision itself will be sampled 500 times so the run will require 100 000 recalculations of the spreadsheet When the run completes DPL displays the Policy Summary The Policy Summary shows that the option to spin off the business is exercised in about 12 of the scenarios For more information on Policy Summaries see Section 3 2 Spin off Gross Margin Yes 13 s1 100 87 does not occur 0 does not occur 0 Trendline Revenue Growth Rate RMC Lawsuit Settlement s1 s1 100 100 does not occur does not occur 0 0 lt gt k Figure 5 38 Policy Summary for the Spin off option 181 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software For continuous chance nodes in the Policy Summary DPL creates two pseudo states s1 and does not occur The probabilities or percentages displayed on these two states indicate how often the chance node occurs in the optimal set of scenarios In Figure 5 38 the percentage for s1 is 100 This means that the chance node occurs in all optimal scenarios This makes sense because the decision tree is symmetric every node occurs on every path In an asymmetric tree See Chapter 6 a continuous chance node might not appear in all optimal scenarios If you d like you can compare this risk profile dataset its expected value and other statistics with one from the model without the downstream decision as des
35. attributes 2 objective 1 Cost1 2 Cost2 constraint 1 Cost1 gt 20000 halt 100000 1 Cost1 2 Cost2 420 Syncopation Software Chapter 15 DPL Programs and Code Because the constraint is only checked before evaluating a node in the tree and not at the endpoints you may also wish to put the constraint expression in the objective function and copy it in the constraint function with an asterisk sequence attributes 2 objective 51 Costl gt 20000 halt 100000 sl Costl 2 Cost2 constraint 15 6 ADVANCED TECHNIQUES FOR PROGRAMS If you do a lot of programming with DPL you may find yourself working with many programs or copying some parts from one program to another DPL offers several advanced techniques for working with your program files that can simplify this activity One technique allows you to break your programs into multiple parts which is especially convenient if you have certain definitions constants for example or a converted spreadsheet model that are used repeatedly Another technique allows you to include several variations in a single program file and choose at compile time which variant should be compiled All of these techniques involve the use of compiler directives which are instructions for the compiler which you embed in a program file If you are using these advanced DPL programming techniques and cannot find answers to your questions in this section or in DPL online Help
36. case the objective function value is 24 7 which is the value at the top of the second bar on the chart and is also the value you will get if you simply run the Wildcat model as usual This value reflects the optimal initial decision alternative which is to Core Sample as well as full flexibility in the downstream option to Drill If there were more decisions in the model DPL would continue from left to right in the tree uncontrolling decisions one at a time and adding a bar for each decision The cumulative height of the last bar in the chart is always equal to the objective function when it is run with all decisions uncontrolled However the starting value for the first bar as well as the lengths of the individual bars may vary depending on your choice of default alternatives Note that the choice of the default alternatives can make a significant difference in the appearance of the chart For each decision option in your model the length of the bar can be interpreted as the incremental value of having flexibility in that option given that all previous options are also flexible uncontrolled and all subsequent options are controlled to the default alternatives you select You may find that you wish to change the default alternatives in the Option Value dialog and save the new settings To do this check the first checkbox in the lower left corner of the Option Value Chart Setup dialog This accomplishes the same thing as changing the
37. frequency histogram eee 88 initial decision alternatives 231 number of intervals eseese 86 reading cumulative ccceeeeee 86 SUAtiSUCS inaa 90 risk tolerance eceeceeeeeeeeeeereenes 365 COCPFICIONE enosn 369 Scenario Risk Profiles nassen 348 sensitivity analysis comparison of types s s s 388 Ypes 375 Sequence section 393 412 SENES aa inie n 253 406 Session LOG aa a aa 7 set keyword seseseseseseseseseserssersnee 417 Spreadsheet IiINKS scseceeeeeeeeeees 437 calculation rn 437 Syncopation Software initialization ceeceeeeeeeeeeeeeeens 437 Manag ng ad cniseeintctanctaaetateora 452 TY DGS oar e 437 stop KEYWOIK ccecceeseceeseeseeeeeaees 416 Strategy table cccccsccecssseeeeeeres 125 Adding decisions to sses 127 Creating strategies 133 Defining strategies 00 135 string KEYWOMK cccseceeeeeeaeeeeeaees 410 subtree DEIFOrMING enaa 227 SUCCESSON evsauniscacnCuntawsaaracanawieccanre 100 SYSTEM requirements 455 time series percentiles 277 481 two way rainbow diagram 375 use tolerance utility ccceeee 418 Utility FUNCH cies iira minenin 372 Value Correlations cccsscsecseeees 110 value keyword ssssssssssssessnnsrenene 403 value node sssssssssssssssrssnsssesrerenne 23 value of perfect information control E EE EN EE AT 107 Value Tornado cseccecseseeeeeees 53 55 color changes ccsce
38. g Figure 9 10 Assumptions Sheet Parameters and Growth Assumptions 264 Syncopation Software Chapter 9 Multidimensional Value Nodes Some of the parameters the ones shaded in blue are already linked to value nodes in DPL The parameters Market 1 Growth Model Choice and Market 2 Growth Model choice are linked so these values which are currently set to 3 are sent from the DPL model If these values are set to 2 a simple annual growth rate the value in cells C9 and or C10 is used for the market size growth If they are set to 3 a different growth rate may be used for each year These rates are currently set in the block of cells that begins with cell E18 The growth model parameters can be set differently by year and for each market In this section of the tutorial you will look at the way the simple annual growth rates are set up in Excel and change the DPL model so that this logic is done in DPL instead using a series in a one dimensional value node Examine the formulas for Market 1 Size Option 2 and Market 2 Size Option 2 these begin in cells E15 and E16 You ll see that the market sizes are initialized using cells that are also linked to value nodes in DPL Market_1_Size and Market_2_ Size After the first year the values Mkt_1_Growth and Mkt_2_Growth which are not currently linked to DPL value nodes are used to grow the market each year In financial and cash flow models as well as in other types of m
39. itialization Mkt_ph2_probs 7 v f Value Liivils A Low Marketing Mkt_ph2_probs phase 2 Nominal 3 20000 ok Cancel _ Hep Figure A 9 Initialization Link for Three State Chance Node Suppose in the example above that the node Marketing phase 2 is conditioned by the 3 state chance node Marketing phase 1 In this case you need to specify 9 probabilities and 9 values for the Marketing phase 2 node since it is conditioned The Data tab for the Marketing phase 2 node would initially look like Figure A 10 449 Appendix A Overview of Spreadsheet Linking Syncopation Software Node Definition Marketing phase 2 General Conditioning Print Page Setup Copy FullScreen Wheel Reorder Multiple Experts a Qq eA Data Distribution type discrete s Links Probability Tiviy Value 8000 Liiviif Marketing 8000 i Low phase 2 Nominal 20000 High 20000 Low Marketing Marketing 8000 phase 1 Nominal phase 2 Nominal 20000 High 20000 Low Marketing 8000 High phase 2 Nominal 20000 High DANAN lt gt Figure A 10 Data Tab for Conditioned Chance Node In Figure A 11 the range Mkt_ph2_cond_probs has been defined and it contains the probabilities for each state of Marketing Phase 2 conditioned on each state of Marketing Phase 1 Note that the probabilities sum to 1 0 across the rows 450 Syncopation Software Kid 9 e
40. the following refers to the value for Sales in 2017 Sales 2017 A relative subscript contains a special identifier that represents the current subscript of the series being initialized series Sales from 2014 0 from 2015 to 2024 Sales 1 10000 406 Syncopation Software Chapter 15 DPL Programs and Code In this case when DPL evaluates the value of Sales for 2015 the symbol stands for 2015 so DPL looks up the value of Sales for 2014 that is Sales 2015 1 and adds 10 000 to it You must initialize each element of a series which means that each interval must be assigned an expression The expression can include the names of variables previously defined in the model and any function Supported by DPL Unlike those of an array described later in this chapter DPL does not allow the elements of a series to be un initialized In addition to using the symbol in the name of a series element you can use the symbol to represent the number of the current index in an expression series Years from 2014 to 2024 Most definitions in DPL can only reference objects defined earlier in a program However DPL allows you to reference a series before defining it That makes it possible to define several series recursively In the following example the definition of Seriesi references Series2 before Series2 is defined series Series2 series Seriesl from 1 1 Erom 2 to 10t SeriresZ 5 L 13 series
41. x86 Syncopation DPL8 Examples folder See Figure 6 1 183 Chapter 6 Asymmetric Trees Syncopation Software E ee DAL Armenen Trees da Symnmetncl EN Homi jade r al a f Dika Thee iL j E l ug D h mitanice de Getacn Conived Ali Lat baal bepee A 1 l Jain hilt ce aa 43 Subinee Uncaninel Ab Aiikdglerance freee y Jf Workipace Manager 4 l a ayy tiig Pepcid Syma Soe dnemmetric with Launch decsion i 3 Session Log 4 n DAL Dect Fiyin quays Copyright E 2008 2008 by lems Tafana br Market vana Pricin hoe peeing sofware of boerae Ie FDA Share Siz cueing Syecrpation amp o fevaie ie Develo E Al rights reagrwred Approval wv wencepgnen sem Low Pa siai Low mpagana US iskren 1566 706 2375 OOOO fo rude tha US a1 978 233 2608 Yes Yes Fi NPV ae m Nominal Nominal Nominal Feise 00 60 No No 3 NPV q Soria F Sy Deas a Sot 4 Leers Ai ae Syoapaties Dofaare P2 High High Y High NPV ia Qe iynmem H De Anei vetth Lunch deti H 46 For hisia preii Fl Iome MON Full Pree l Figure 6 1 Simple Model with a Symmetric Tree This DPL model and its associated spreadsheet AsymmetricTrees xls are set up to calculate the value of a pharmaceutical R amp D project The project is in an early stage so there is uncertainty about whether it will succeed and result in a drug with regulatory approval
42. 1 value p 55 value g 0 excel Excel 1 Assumptions Market 1 Size 394 Syncopation Software Chapter 15 DPL Programs and Code value Market 1 S1ize 110 excel excel 1 Assumptions Market 2 9126 value Market 2 Size 180 excel Excel 1 Assumpt1ons Mkt 1 Model value Market 1 Growth Model 3 excel Excel 1 Assumptions Mkt 2 Model value Market 2 Growth Model 3 excel Excel 1 Assumptions Proceed decision Proceed Yes No dee Proceed Yes 07 Proceed No excel Excel 1 Assumptions Initial Upstream Investment decision Initial Upstream Investment Low gt High 620 Initial Upstream Investment Low 630 Initial Upstream Investment High decision Pipeline Stravegy Mkt l 10 BEM Mkt I 20 BOM Mkt 1 and 2 10 BCM Mkt 1 and 2 20 BCM Mkt 1 and 2 30 BCM excel Excel _ 1 Assumptions Include Market 2 decision Include Market 2 Yes No 2500 Include Market 2 Yes 0 Include Market 2 No excel Excel 1 Assumptions Pipeline Capacity decision Pipeline Capacity BCM_10 BCM 20 BCM 30 LOY Pipeline Capacity BCM 10 20 Pipeline Capacity BCM 20 30 Pipeline Capacity BCM 30 chance CapEx Low Nominal High 2 4 4 excel Excel 1 Assumptions CapEx value CapEx Pipeline Capacity CapEx Pipeline Capacity BCM 10 650 Ei CapEx Low 1000 Ea CapEx Nominal L750 ley CapEx High Pipeline Capacity BCM 20 T50 yy CapEx
43. All Objective Function Use defaults based on Policy Tree selection don t show dialog again Can be reset in Home Run Options cence Figure 12 10 Scenario Risk Profile Dialog with Market Size High gt Click OK DPL generates a new Scenario Risk Profile for this scenario Note the title of the Risk Profile indicates what the scenario is Also note that this Scenario Risk Profile replaced the previous one since you did not rename the previous one Finally note that this Scenario Risk Profile corresponds to a set of endpoints that cannot be defined just by starting from a location on the Policy Tree since the three events whose states are fixed are not at the head of the Policy Tree and consecutive The Scenario Risk Profile dialog has a checkbox option to use the default settings for each event based on where you are in the Policy Tree If you do not want to be prompted with the Scenario Risk Profile dialog check this option When you select Scenario Risk Profile Add the Scenario Risk Profile will automatically be created using your selection in the Policy Tree You can also change this setting in the Home Run Options Note that if this option is checked you cannot create a Scenario Risk Profile similar to the one you just created since there is no point in the tree that matches the selections you made in the dialog Close the Workspace without saving changes You will not ne
44. Asa to ws ampie Da Rename 2 PP Frame iJ bjectiwe F India 030 Ke t e LN Workspace Manager a S Learning gt Wikkat simple Session Log 9 OPL Decision Pr Ter ge opyright 2003 ncorperatrg sole Driling Amount Qil yreopaion Softw Costs of Ol Poe Al ngit ved WW suncogyien Gore Drill M as supron sympatien con Low y Low E Diling Costs Revenues ae Yes G Med Wet Normal 1 Profes j Dning Costs A 4 Revenues Perset j High Soaking High Release 2 00 00 4 Senai Member SYOPS013010200 E Drilling Costs Revenues Leoereed User ober Syncopeton Sofware he 1 pang 11 2222 No q p gt Widcat simple x a gt Fe t Heip press FI Zoorm tw Full Prev Gs Figure 7 1 Simple Wildcat Decision Model 216 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models This model represents the decision problem faced by a Wildcatter exploring for oil In a given field s he must decide whether or not to drill If oil is found uncertain costs are incurred If a dry hole is drilled nothing is gained If oil is struck then an amount of revenue is gained which depends on the price of oil and the reserves found Inthe Home Run group check Risk Profile and Policy Tree Uncheck the Init Dec Alts checkbox Figure 7 2 gt Click Home Run Decision Analysis Risk Profile Policy Tree Levels S all Hi Objective Function ol Policy Summary C Endpoints Decision
45. Case Tornado no policy changes will be seen in this model However for models with subsequent decisions after the initial decision often referred to as downstream decisions policy changes may occur in an Initial Decision Alternatives Tornado since the optimal alternative may change for a downstream decision You may want to save the workspace before you proceed to Chapter 3 Decision Analysis Results 2 Syncopation Software Chapter 3 Decision Analysis Results 3 ANALYZING DECISION ANALYSIS MODELS The model you built in Chapter 2 is now ready to produce the full suite of results that are available within the Home Run group when you run a decision analysis We will discuss each of these options in more detail now See Tables 3 1 through 3 3 below TI Ey This section determines how DPL will analyze the decision Decision tree Analysis Fast sequence evaluation Provides the exact expected value of the objective function but optimizes run time by using the structure of the tree to avoid visiting every endpoint of the tree The Risk Profile and hence percentiles may be approximate Full tree enumeration Provides exact expected value of the a objective function and visits every Y E endpoint of the tree The Risk Profile eae and percentiles is exact and the Endpoint Database can optionally be recorded for re use Discrete tree simulation Provides an estimate of expected value by sending a specified number of sa
46. Case Tornado Diagrams are useful for understanding how the chance nodes in your model impact the objective function They are also flexible because you can add bars to a Base Case Tornado Diagram for values in your model This allows you to compare the impact of chance nodes across their range of states to the impact of changes in values across a range of settings Lastly Base Case Tornado Diagrams are very quick to run which may be helpful for larger models with many chance nodes While Base Case Tornado Diagrams are relatively simple and easy to understand they can be limited in terms of showing impacts in a probabilistic model where there is a lot of conditioning and or where probabilities represent skewed distributions Base Case Tornado Diagrams don t use probabilities so if for example the probabilities of a chance node reflect a skewed distribution the Base Case Tornado will show the same results as for a symmetric distribution A Base Case Tornado is also difficult to use when there is no natural middle case as in a risk analysis model which considers a number of very unlikely bad outcomes In later chapters we ll discuss more advanced tornado diagrams that can be insightful for complex models See Section 14 3 When to Use Which DPL Sensitivity Output for a comparison 69 Chapter 2 Building Decision Analysis Models Syncopation Software 2 09 INITIAL DECISION ALTERNATIVES TORNADO DIAGRAM The Initial Decision Alternatives To
47. Chapter 8 Objective Function and Multiple Attributes Objective amp Utility Wildcat Objective Attributes Risk Tolerance insert Delete V Gather attribute expected values in the Policy Tree Objective function Maximize Minimize Units for output labels Constraint function Roll forward 6 Roll back 2it Land lt 10 Profit Land_Weight Land halt 1e12 B lv fa Figure 8 16 Objective amp Utility dialog with Constraint Function When the halt function is evaluated DPL will stop evaluating the current decision tree path and will insert an expected value of negative 1e12 In other words it would cost you one million trillion to violate the constraint so the Land attribute will definitely never be greater than 10 gt Click OK Notice that the new objective function is displayed in the Objective Function combo box in the Model Objective amp Utility group In addition to the Objective amp Utility dialog you can edit the objective function using the Model Objective amp Utility Objective Function combo box either by selecting a single attribute from the drop down list or by typing an expression involving more than one attribute and or value from the model into the edit box The select attribute value and function buttons are also available on the ribbon gt Click Home Run Decision Analysis gt Click OK to the warning 251 Chapter 8 Objective Functions and Multiple
48. DPL does not even explore the other alternatives to see if they are optimal The only way you can include a controlled event in the sequence section of a program is by setting its state 417 Chapter 15 DPL Programs and Code Syncopation Software You can only set the state of a decision or chance event if it has not occurred previously along that path through the Decision Tree Also once you set the state of a decision or chance event you cannot use keywords like decide or gamble to make the event occur at a subsequent point along the same path In contrast you can set the state of a controlled event several times along any path decide to Advertise Yes and pay Ad Cost then set Demand High and get Profit to Advertise No then gamble on Demand and get Profit 15 5 9 Using a Utility Function to Describe Risk Attitude The expression for DPL s built in exponential utility function is exp V T where V is a value and T is the risk tolerance coefficient To use it you only need to specify the decision maker s risk tolerance coefficient value RiskTolerance 100 sequence use tolerance RiskTolerance and decide EO waa The expression for the risk tolerance coefficient cannot contain values that depend on events Also this expression cannot contain the state function Because the specification occurred at the beginning of the sequence section it applies to the entire tree It is possible albeit unusual to specify different risk
49. Develop Nominal Yes PV_Development No Figure 6 25 Influence Diagram and Decision Tree with Key competitor outcome Node The previous model assumed you would launch the product if you obtained FDA Approval You ll now add an explicit launch decision Drop down the Model Node Add split button on the ribbon see Table 1 2 and choose Decision node from the list The mouse cursor changes to a semi transparent decision node Click in the Influence Diagram The Node Definition dialog appears Name the decision Launch Leave the default alternatives Yes and No as S Click OK Rearrange the Influence Diagram so it looks like Figure 6 26 You ll need to remove and add several influence arcs You can optionally bend the arcs as they are in the figure as well Y Yd 205 Chapter 6 Asymmetric Trees Syncopation Software Key Market Market competitor Share size aan outcome Bae mae Launch Approval Low Low Low Failure Nominal PV_Launch Yes Nominal Nominal Disappointment PV_Market lt 1 No High High Breakthrough PV_Market Figure 6 26 Influence Diagram with Launch Decision PV_Development No In the spreadsheet the Key competitor outcome influences the size of the market for the drug under consideration if the competitor has a breakthrough the market for the drug being considered for development will shrink Initially al
50. Do File Save as to save the Workspace with a new name If you would like to run a Time Series Percentiles now skip to Chapter 1 If you would rather learn more about multidimensional value nodes first continue to the next tutorial 9 1 2 A One Dimensional Series Export Node This tutorial provides an example of a one dimensional value node containing a series A series in DPL is a one dimensional vector of values that may depend on the states of events may use formulas or may just contain constants Series are different from one dimensional arrays because they are defined using intervals and or relative subscripts as will 263 Chapter 9 Multidimensional Value Node Syncopation Software be demonstrated in this tutorial The ability to use relative subscripts in a series makes it easy to define elements of the series in terms of other elements e g sales in time period t is equal to sales in time period t 1 times a growth factor Either continue with the example you saved in Section 9 1 1 above or if you closed them re open Multidimensional Value da and Multidimensional Value xls Note If you completed the Time Series Percentiles tutorial in Chapter 1 and saved your model you may use that model for the remaining tutorials in this chapter however your model will look slightly different from the figures in the following sections In Excel select the Assumptions tab of the spreadsheet and look at the Basic Paramete
51. Endpoint Database Without View in Workspace Manager You can restore the view of the Endpoint Database by double clicking on the icon in the Workspace Manager You can also play endpoints without them being displayed in a report as you will see in the next section If you select the Endpoint Database item at this point and press Delete then the Endpoint Database itself will be deleted 11 2 5 Risk Profiles On occasion you will want to play endpoints for the purpose of generating output charts that you did not generate with the initial decision analysis run You can generate four types of outputs by simply playing the endpoints Risk Profiles Policy Trees Policy Summaries and Expected Value of Perfect Information and Control charts If your model has multiple attributes you may want to play endpoints in order to generate Risk Profiles for attributes other than the objective function See Chapters 8 1 of this manual for more information on models with multiple attributes 323 Chapter 11 Using the Endpoint Database Syncopation Software You can also play endpoints to generate an Initial Decision Alternatives Risk Profile You will do this now Make sure Risk Profile and Init Dec Alts is checked gt Also check Policy Tree and Policy Summary for comparison with the results of the next sub section of this tutorial Run a Full Tree Enumeration from Endpoints DPL quickly creates an Initial D
52. Examples Navigator icon Contextual Tabs and Commands Some ribbon tabs are contextual and will appear only when a particular item is active For example the Policy tab appears only when a Policy Tree or Policy Summary is active As you work your way through the following tutorials in addition to the Policy Tab you will find the following contextual tabs the Grid Tab when an Endpoint Database is active and the Chart Series and Chart Format tabs when a chart such as a Risk Profile Chart or Value Tornado Diagram is active 19 Chapter 1 Introduction Syncopation Software Furthermore not all commands within a particular tab will always be enabled since some of them are context specific For example the options within the Model Get Pay group are only enabled when you have a branch selected in the Decision Tree 1 4 7 DPL and Workspace Level and Model Options You will now check that several options are set in specific ways so that as you work with DPL what you see on your screen will be similar to the images in this manual In DPL 8 there are several places to set options You can access DPL and Workspace level options via File Options note in this case File Options is a menu item not a dialog box launcher Options that are specific to models within a Workspace Model Options are set via dialog box launchers that relate to the options being set e g options relating to nodes are set via the Model Node di
53. Formula Formula Formula Years _to_peak Figure 6 16 Range Names Dialog gt Click OK Initially DPL will place the nodes to the right of NPV you may have to scroll the Influence Diagram pane to the right to see them Move the nodes so that they are positioned as in Figure 6 17 Add two new arcs as shown in Figure 6 17 197 Chapter 6 Asymmetric Trees Syncopation Software Figure 6 17 Influence Diagram with Three Import Nodes You also need to change the NPV node so it includes only revenues and costs associated with a product on the market gt Double click on NPV in the Influence Diagram The Node Definition dialog appears Click on the General tab Change the node name to PV Market Click on the Links tab Click the Cell Names button Select PV_post_launch from the list Click Select The Node definition dialog is updated Figure 6 18 UGEUY 198 Syncopation Software Chapter 6 Asymmetric Trees z Node Definition PV Market General Data Calculation links These links are recalculated throughout the run They may be either inputs used in the spreadsheets calculations e g unit sales or calculated results from None local DPL Program Microsoft Excel Workbook C Program Files amp 86 Syncopation DPL8 Examples Asymmetric Trees xs Browse Sheet Cell DCFIPV_post_launch Go to Cell y This is an import node because it doesnt have data Its va
54. High Soaking High Drilling_Costs 0 Seismic Structure Core sample Test_Costs 0 Figure 8 15 Decision Tree with Get Pay Indicator 249 Chapter 8 Objective Functions and Multiple Attributes Syncopation Software 3S 2 USING A CONSTRAINT FUNCTION The objective function gives you considerable flexibility in defining how much you like dislike each attribute but there may still be situations where you want to place an absolute limit on one or more attributes In DPL you do this by specifying a constraint function A constraint function is normally an if statement with a halt function for its False value The halt function prevents DPL from continuing down that path This serves to limit the size of the tree as well as enforce the constraint The syntax for the if function is if condition True value False value Navigate to the active Model gt Click Model Objective amp Utility Options The Objective tab of the Objective amp Utility dialog will appear gt Click in the Constraint function edit box Type if Land lt 10 Profit Land_Weight Land halt 1e12 DPL is case sensitive so make sure your spelling capitalization and punctuation are correct When a constraint function is used the objective function and the constraint function will normally be the same gt Copy the constraint function up to the objective function edit box See Figure 8 16 250 Syncopation Software
55. Mkt_1_Growth Links ox Gees ta Figure 9 13 Node Definition Data for Mkt 1 All Years Node Take a moment to understand how this series is defined The series definition refers to itself using the relative subscript symbol The term 1 indicates the previous element in the series In DPL series are the only type of expression in which you can use relative subscripts See Chapter 15 of this manual for more information This series is now defined to produce exactly the same result that the Excel formulas for Market 1 Size option 2 are producing The series is indexed according to the years in the cash flow model 2014 through 2038 This was done for clarity of definition For purposes of the analysis you could also have indexed the series from 0 to 24 or from any other integer to that integer plus 24 DPL will only export the actual values of the series expressions to Excel 268 Syncopation Software Chapter 9 Multidimensional Value Nodes Now you need to link this node to the appropriate range in Excel so that DPL can export values There is already a named range in Excel to do this LUULU If you d like to see the named range switch back to Excel Select cells E15 through AC15 Look in the Name combo box in Excel and note it says Mkt_1_All_ Years Note there is no requirement that you name this range exactly the same as the value node in Excel Any range name acceptable to Excel is
56. Model Often you will build a model in DPL and a separate spreadsheet in parallel and link the two at some point in the process adding new nodes to DPL that are linked to the spreadsheet In this situation you may wish to use the command Model Link Excel Calculation linked 439 Appendix A Overview of Spreadsheet Linking Syncopation Software This command adds linked values to the Influence Diagram The nodes are created as values but as with Method 1 they can be changed later to decision nodes or chance nodes as desired The basic steps for this method are as follows Select Model Link Excel Calculation linked If you have not created any Excel linked nodes browse to locate the spreadsheet you wish to use otherwise DPL uses the currently linked spreadsheet Select the names of the cells for which you wish to add linked values See Figure A 3 Range Names C Program Files x86 Syncopation DPL Examples Cash Flow with Decision xls Creating Calculation Linked Yalues Mame Consulting_phase_1_ growth Coneulting phase growth Discount_rate Initial consulting Initial license lritial training License phase 1_ growth License_phase_ _ growth Marketing _phase_1 Marketing _phase_2 Mkt phl_probs Mkt phl_vales Mkt ph _cond_probs Mkt ph _probs MktLphe_ values NPY Payment fee Pursue_Consulting TOTAL _COST_OF_SALES TOTAL_OVERHEADS Training _phase_1_ growth Training phase
57. Nominal Sion Actual Price High Comments such as the above are not necessary but they usually make it easier to read and interpret the DPL program The last elements in the definition section are the definitions of the two Excel linked value nodes that import the cash flow results excel Excel 1 Assumptions Total NPV value Total NPV excel Excel 1 Assumptions Uptiront NPV value Upfront NPV Scroll down until you see the line that says sequence The section of the program after the line sequence represents the Decision Tree This section tells DPL the order in which it should evaluate the nodes and which Get Pay expressions to evaluate during the analysis 397 Chapter 15 DPL Programs and Code Syncopation Software Notice that since the first four events on the tree are symmetric the DPL code for this portion of the tree is very simple decide to Initial Upstream Investment then gamble on Appraisal Results then Gamble orn Forecast Price then gambl on Market 2 Political Stability Now them decides Following this section the DPL program describes the asymmetric portion of the tree In this last part of the program the indenting of the program text matters See Section 15 5 for detailed information on the sequence section If the Program Window is active in the Model Window DPL will analyze the program when you run an analysis gt With the Program Window active run the program as you n
58. Note also that Saved Endpoints is still saved as a separate item 329 Chapter 11 Using the Endpoint Database Syncopation Software Workspace Manager Sr Pro_Ch6_Fig6_17 da Me xe Energy Model S Endpoint Database z Saved Endpoints oll y Policy Summary lv Policy Tree Figure 11 21 Workspace Manager after Reconnecting gt Check Risk Profile Init Dec Alts Policy Tree Policy Summary and VOIC Select Home Run Decision Analysis Full Tree Enumeration from Endpoints DPL instantly creates new outputs from the original Endpoint Database You can examine the Policy Tree or session log to see that the expected value is back to 34 9 as before In addition DPL has generated a new Value of Information and Control chart from the reconnected endpoints Examine this chart if you like Value of Information and Control charts are explained in the Chapter Section 3 4 of this manual You do not need to save your model for use in the last two sections of this chapter If you do save the model in its current state it will be a large file because there are two Endpoint Databases one of which is displayed in a report in the file Right Click on the Saved Endpoints item and select Delete The Endpoint Database in this report was not connected to a model so it will be deleted entirely Close the Excel file Energy Case xls without saving changes 11 4 EXPORTING ENDPOINTS As you h
59. P Low Le L Eshi Low HTI a Ya Low hia He BCH Le Low Le irabis 4 igh mii Wee Liw Wal He BOM Le Lew Low Suk G J att Vt Live Wiii Tee He BCh e Low Leow bow Shab A g tgh 10 Wet Live hia He BCH Low Low Ls Eik Pisma Chk TI Wee nV Wir l Ha BOMI Low Lew Low Surbir fort att l Wee L Aiti No BCh Low Low Low Stable homes gh 15 Wen il hiit Ho ECHI Low Los bows Teubla Morr m oh 4 14 Wes Mia Ne BCM II Low Lew Leow Sub wen om 15 Wes tihe BEMI Le Low Le ina hore qn Feo Log p Vet a He iM is Le be Seubbe Mom ig aji DFL Denson Pop emng LIH Tti Wiin H BOMI L L L Seabee t Ha ba i patio pare fic ae i Wie hit 1 Ne BEM i L L f rable Pamen Higi ah vi saa ny ail E TEL hiti BCh L t Sable bi oh _ Ign neers Taf hii F BCM Lin L L Sabie Hix Gi ior I ts oS a BOM Low L L rhe Hg 8 Hegh oi smm a 7 ie hit DCM i Le Tbs hig imina bpw cha 1 TH 3 i ht l BCH E Lon Le Le Stable hir ic mank bm r biin i Boh Lo Le Lew Sake H hammi begh bitit Mid bce Le Le i Seabbe sagh sk pions te bi a khia BCH l l rakes 7 igi arai h j ai BCH L Low Lew Saabs H Hgh Ha TAA EA E A 4 i iit T m i iM i be i inertia mm J hiia y BCM i l l sable Spat 7 i i BCM L Le L bir P Erie fi b ae nany Medi H ndpnrni Detubhace N gf pria via w Polry Trg W ab Figure 11 2 Endpoint Database Report The first column in the Endpoint Database indicates the endpoint number Following that the Endpoint Database has a column for each event decision chance or controlled
60. Percentiles continuing the example from Section 9 1 1 which uses a 1 dimensional value node that has been previously defined and linked to Excel Time Series Percentiles require the use of DPL s multiple attributes feature as described in Chapter 8 of this manual This chapter also covers how to quickly generate Risk Profiles for multiple attributes simultaneously as well as more information on how to use the Get Pay tab of the Branch Definition dialog with multiple attributes Note that these features can be used in any multi attribute DPL model not just models in which the attributes correspond to metrics over time 10 1 TIME SERIES PERCENTILES Select File Open Navigate to the Examples folder underneath where you installed DPL If you used the default location the path is C Program Files x86 Syncopation DPL8 Examples Select Time Series Percentiles da DPL opens the Workspace as shown in Figure 10 1 277 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software aal nn i GW DEL Time Senes Percenbles Energy Model aed Fic hine Bineli EE Cote Heip Agd fe WE Delete Setect All Ti Pie Predile 7 Poy Tree Levels 14 jn Monte Carts Surg Tanned Rename Jg copy J MindFepiee Mar Cbjectie Function Policy Summary y Eredpoinis Ingial 100 7 Comm pil Paste Me aotanode iniji F irit Det Ata vai Vabar Comelabiens Restart a0 hy Fainby OF ake a Workigace Manyi a Tame bem
61. Profit Click OK Type a plus sign Click the Variable button x Select Land_Weight Click OK Type an asterisk for multiplication Click the Attribute button tsh Select Land Click OK Check the checkbox next to Gather attribute expected values in the Policy Tree Now when you run your model and produce a Policy Tree the rolled back expected values of all attributes will be displayed in the tree This option is referred to as Fat Policy The Objective dialog should look like Figure 8 7 241 Chapter 8 Objective Functions and Multiple Attributes Syncopation Software Objective amp Utility Wildcat Objective Attributes Risk Tolerance Gather attribute expected values in the Policy Tree Objective function Madmize Minimize Units for output labels Constraint function Roll forward Roll back is vf Figure 8 7 Objective amp Utility Dialog gt Click OK to close the dialog You may wonder why you didn t use the value Land Disturbed in the objective function Land Disturbed will contribute to the Land attribute in a Get Pay expression in the tree Separating attributes and values gives you greater modeling flexibility For example you might later have several value nodes which all contribute to Land just as you already have three value nodes that contribute to Profit Test Costs Drilling Costs and Revenues You ll now add a Get Pay expression for Land D
62. RISK TOl ranCe eaoaai aA 365 13 1 Incorporating a Risk Tolerance ssssssssssssnsrnsrrsrrnnrsnnrsnrnnrsnrnnrnnrnnrnnrnnennnne 365 13 2 Advanced Utility Functions wissicsssxincinicsmintnicsaawine saneutun ube riadabxdasnaneamalembvenbes 372 14 Sensitivity Analyses cccscscscscseeeeeeeeesesseeesenenenenensnenseessseeeseeeeeenes 375 14 1 Two Way Rainbow Diagrams ssssssssssssssrsrrsnrsnrnnrnnrrnnrnnrsnrnnennrnnrnnnnnennnne 375 14 2 EVENE T OFM AGOCS srd a ae aaa a Erina EATA TATAA 384 14 3 When to Use Which DPL Sensitivity Output s sesssssssenssnsrnnrsrrnrenrenrnnenne 388 15 DPL Programs and Code ccscscssssssssseesesesesenenenensnsnsescssssnseeeenenenes 391 15 1 Whats a DPE ProGrain savsdusvcusosidnsicneinonivnniwnsadnsivnsaveeisesieSaniendencsd nes sesatt 391 15 2 Overview of DPL Program Components ss sssssssssrssrenrrsrrnnrnrnnennnnnnnrenrennne 392 15 3 Converting a Model to a ProGraM cccccseeceeeeeeeeeeeeeeeseeesaeesaeesaeesaeeeaeeeaees 394 15 4 Defining Components in the Definition Section sssssssssssssrsrrsrrsrrerrenrsenns 399 15 5 Defining Event Sequences in the Sequence Section sssesssssrrersrrsrrsrsnee 412 15 6 Advanced Techniques for ProgramS ss ssssssssssssrssrrsrrsrrsrennrnnrenrenrnnrenrennns 421 16 Converting Spreadsheets to DPL Code sssssssunnnnnnnnnnnnnnnnnnnnnnnnn 425 16 1 How to Convert SpreadsSheets cccsccsccssecseeseeeeeeeeeseeeseeseeeeeeseeeeeesseeesaeess 4
63. Samples JP Sample Se Companson z gt Dei Aa l ion Log 9 GePL ema tiraia L TY ie i irri penere E eiu on ioenas E A SH ave ri Screamer in Hyi hii ipiri EH a EEA Pa iiet ss reer UL Tete com US intree 1 B86 oie Cutuide tre US 1 F share of at Gr Irori BML Risk Preteen Vereen Maan Revenue Lawsuit Re rll Perpara Ieee i eee P nan Growth Rate Rate pie a niet SAO oo ht eae let aot Syncopalon Solera Fi Total alin Gere a eo pine anak nasa t arial narmal Caa E ular EP Mite Carte LL ld dale Way 9 2008 z me Comprighd DIS by 5 sepatien Safwan re a iie a Ak Anahi A Temeda X ff irele Sr Compersien R a ec ahii Jp Pm pret F ie E Fall P Figure 5 25 New Model for Decision Analysis 169 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Assume you ve prepared a spreadsheet which includes two new input drivers one for the spinoff decision and one for its effect on the growth rate You ll need to link the decision analysis model to the new spreadsheet Select Model Links Options The Model Links dialog is displayed see Figure 5 26 Click the Browse button Select Enlightify2 xls Click Open If you can t see the full path and name of the newly linked spreadsheet Enlightify2 drag the Name column separator to the right You should see that Enlightify2 xls is now linked to the model Click OK GUU 9 4 Mode
64. Select the branches of the Launch decision and press Ctrl X The expression is placed on the clipboard Select the branches of the Pricing node and press Ctrl V The Get Pay expression is removed from the branches of Launch and added to the branches of the Pricing node You are now ready to reorder the tree so that the downstream decision is made after learning about the outcome of a key competitor gt In the decision tree drag the Key competitor outcome node and drop it on top of the Market Share node The Key competitor outcome node is now connected to the Yes branch of the FDA Approval node and comes before the Market Share node See Figure 6 27 Market Market a Pricing FDA Approval bow Low Liat Develop E n i T f F PV Market _ Market Yes Normal Nominal Nominal Nominal g Teg u Launch gt e e sate Market No of PV_Development _No 4 High _ High High 1 lt Ae lt q sd _ Market Figure 6 27 Decision Tree with Key Competitor outcome node moved 207 Chapter 6 Asymmetric Trees Syncopation Software Drag the Launch decision node and drop it directly on the Market Share node Your decision tree should look like Figure 6 28 Key competitor Market Market outcome lamh Share Sve Pricing FDA Approval Low Low Low Low velop paaa Develop Yes j PV_Market Yes Nominal Oo Nominal Nominal Nominal lt q yes PV_Launch _No 9 M PV_Market SV Development a No High
65. SetA SetB SetC Spend Eiswusqu Addfeatures Addfeatures Addfeatures Increase Increase Increase Increase eee a ee a ee Keep same Keep same Keep same Keep Same Keep same Keep same Keep same Minimal Minimal Minimal Decrease Decrease Decrease Decrease Figure 4 10 Renamed Status Quo Strategy 132 Syncopation Software Chapter 4 Strategy Tables You may also double click an item to rename it Rename Strategy2 to Feature Led You will now create two new strategies Click the Commands button and select Add Strategy from the list as shown in Figure 4 11 Node Definition Strategy General Full Screen Copy Format Full Screen Data Add Strategy p Feature Feature Marketing Price Price Price Change Change Change Add New Decision Set B SetC Spend A z c Add Alternative es Addfeatures Add features Increase Increase Increase Increase Add Existing Decision Pp Break Out Decision Rename Delete e Keep same Keep same Keep Same Keep same Keep same Keep same ere Minimal Minimal Minimal Decrease Decrease Decrease Decrease Figure 4 11 Selecting Add Strategy DPL creates a new strategy called Strategy3 color coded blue and puts you into rename mode as shown in Figure 4 12 133 Chapter 4 Strategy Tables Syncopation Software Node Definition Strategy General Commands Print J Page Se
66. Software 7 2 SUMMARY OF DPL INFLUENCE ARCS DPL influence arcs are color coded Table 7 2 summarizes what each color 3 D V S YW Black gt No conditioning Blue Values conditioned Green Probabilities conditioned Light orange Both values and probabilities conditioned Table 7 2 Arc Color Meanings Arcs appear magenta when selected so you may need to press Esc to see the actual arc color The from node is known as the predecessor and the to node is called the successor Because not all predecessors have states and not all successors have probabilities some arc types may not be available between a given pair of nodes Even when an arc is black and there is no conditioning there can still be dependence in the calculations if the value of the predecessor node is used in a formula in the successor node This is the case with the arc from Oil Price to Revenues in the Learning da model 232 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models Table 7 3 summarizes which type of arc is available between each of the node types in DPL To Successor chance chance predecessor chance Discrete Black Black Black chance light green blue orange blue light orange Decision Black Black light blue orange Controlled Black Black light blue orange Value Controlled Black Black blue blue Black Black blue blue Black Black Black blue blue
67. Tree to make it active Select the chance node for Key competitor outcome You may need to scroll left slightly to see it gt Click Policy Scenario Risk Profile Add You will see the Scenario Risk Profile dialog as in Figure 12 8 348 Syncopation Software Chapter 12 Advanced Decision Analysis Results Scenaro Risk Profile Drug Development Develop_ FDA_Approval Key_competitor_outcome Launch Market Share Market_Size Pricing Objective Function C Use defaults based on Policy Tree selection don t show dialog again Can be reset in Home Run Options Figure 12 8 Scenario Risk Profile Dialog You asked for a Scenario Risk Profile from the point in the Policy Tree where the initial decision Develop is Yes and the outcome of FDA Approval is Yes The dialog is initialized to these settings and indicates that this is the scenario you have selected All other events in the tree are denoted All which means that they are allowed to vary as usual only the initial decision and the first chance node are fixed gt Click OK to generate the Scenario Risk Profile See Figure 12 9 349 Chapter 12 Advanced Decision Analysis Results Syncopation Software O Jd C OPL Poticy Toreda Scenario Expected Yaksa ES rie Home Medel ver Dots twe Format Hee LaF Wow Pd a Aegre Vojapace Mansger 3 FDA Approval Yes Develop Yes 100 Curnulabwe Probability
68. Ty epo 9 TO tag SZ To refer to the element in the first row and first column of array B you use the syntax A 0 0 which equals 1 The element in the first row and second columns is A O 1 which equals 2 and so on Commas separate elements in a row in two dimensional array Semicolons separate rows in a two dimensional array Note the last row does not end with a semicolon The elements of an array unlike those of a series cannot be initialized in groups You must initialize each element of an array separately although you can leave some elements un initialized If you do not initialize an array element it is called a null element The subscripts of an array are always in a range from 0 to a positive integer If necessary DPL will truncate the number represented by an array subscript to produce an integer 408 Syncopation Software Chapter 15 DPL Programs and Code If you would like to declare an array that has multiple rows and one column you must do this as a two dimensional array The array Payments below is an array with four rows and one column array Payments 100 150 200 240 To access the third row of this array you use Payments 2 0 You may not omit the 0 Note one dimensional arrays are treated like two dimensional arrays with one row Therefore you can refer to the elements of array A declared above as either A O 1 or A 1 The 0 is optional You can initialize each element of an array wi
69. Way Rainbow Diagram you must have Show Tips turned on To check if you have Show Tips turned on go to Files Options General gt Click OK to close the Options dialog gt Move your mouse over one of the markers in the Two Way Rainbow Diagram 381 Chapter 14 Sensitivity Analyses Syncopation Software A policy tip appears as shown in Figure 14 8 Price Change Raise 30 Keep same 0 Product Strategy In house 100 Figure 14 8 Two Way Rainbow Diagram with Policy Tip The policy tip tells you the expected value of the model given the values of the two variables that the marker represents For example the policy tip in Figure 14 8 indicates that the expected value of the model is 33 9 when License_Fee is 0 825 and Per_Unit_Cost High is 0 875 The policy tip also tells you the policy dependent probabilities for each decision in the model given the values of the variables that the marker represents The decision alternatives are displayed in the policy tip by using the decision node name followed by a colon followed by the decision alternative E g Product Strategy License The policy dependent probability for the alternative follows the equal sign As Figure 14 8 indicates the policy when License_Fee is 0 825 and Per_Unit_Cost High is 0 875 is to produce the product in house and downstream to raise the price in 30 of the scenarios and to keep price the same in 70 of the scenarios To keep the policy tip compact
70. We Raise car No on Nominal Nominal Nomina d License c Keep Same ny d High High High LLL Figure 3 23 Model with Total Unit Sales Now we need to update Revenue_per_Unit to reflect the price change The formula for Revenue_per_Unit in the In house case becomes an average selling price Double click Revenue per Unit to edit it gt The In house formula is Price Sales_1 Price Price_Change Sales_2 Total_Unit_Sales The License formula doesn t change Again use the Variable button for accuracy See Figure 3 24 If the data is cut off in the input tree you can right mouse click on the tree to get the context menu and select Lengthen Branch 102 Syncopation Software Chapter 3 Decision Analysis Results Node Definition Revenue per Unit be K P Fe General Conditioning Print Page Setup Copy Full Screen E czars wa Revenue In house per Unit Price Sales_1 Price Price_Change Sales_2 Total_Unit_Sales Product Strategy License License_Fee O a Hep Figure 3 24 Node Definition Dialog for Revenue per Unit gt Click OK close the dialog Add arcs from Sales 1 Sales 2 Price Change and Total Unit Sales to Revenue per Unit You have now incorporated Sales 2 and Price Change into the expressions for Revenue per Unit In particular if In house is chosen then revenues reflect the price change in the second period Your model is near
71. a number when you add an attribute This makes it easy to add sequential attributes as in this example If you did not want a sequentially named attribute you can always rename it Now that there are multiple attributes defined in the model you need to tell DPL explicitly what the objective function is If you do not specify an objective function when there are multiple attributes defined DPL will maximize the sum of the attributes by default You do not need the ten new attributes for the objective function The objective function is still simply NPV Click in the Objective function edit box Type NPV or use the select attribute button S to select NPV for the objective function 282 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics The maximize radio button is already selected DPL will now maximize NPV for the objective function The Objective amp Utility dialog should look like Figure 10 6 Objective amp Utility Energy Model Objective Attributes Risk Tolerance C Gather attribute expected values in the Policy Tree Objective function Maximize Minimize a A Units for output labels Constraint function Roll forward 0 Roll back s w ly Figure 10 6 Objective amp Utility dialog after Edits Have Been Made Click OK The last step to set up Time Series Percentiles is to change the Get Pay expressions in the decision tree to tell DPL which valu
72. amount of uncertainty in the model called the Original Model Range Then DPL runs the model one additional time for each chance event in the model the sensitivity event However for Probabilistic Event Tornadoes in these subsequent runs DPL leaves all of the chance events in the model unchanged except for the sensitivity event The sensitivity event is replaced with its expected value In effect in each sensitivity event model run the reduction in uncertainty from the Original Model Range is due to the removal of the uncertainty associated with the sensitivity event If either Don t Gamble or Always Gamble is set for a chance event this has an impact on how the Event Tornado is run The Don t Gamble and Always Gamble checkboxes are found on the Tree Instance tab of the Node Definition dialog To access the Tree Instance tab for a chance node double click the chance node in the Decision Tree or select Model Node Edit while the node is selected in the Decision Tree For Probabilistic and Deterministic Event Tornadoes Don t Gamble and Always gamble have the following effects If Don t Gamble is set for a chance event then it is not gambled on regardless of whether or not it is the sensitivity event i e it is always replaced with its expected value in the Event Tornado If Always Gamble is set for a chance event then it is 384 Syncopation Software Chapter 14 Sensitivity Analyses gambled on regardless of whether or not it is
73. amp amp Use the drop down list below the Add label to add Actual_ Price Col_7 to the filter Place your cursor after the Col_7 in the Custom filter edit box and type lt 2 The filter should now read Geot 3 S40 i SS SS Dy 6s Col 8 SSO Se Cor lt 2 You have added an AND condition such that Col_7 is less than 2 This last condition corresponds to Actual_Price being Low 0 or Nominal 1 Note you could have created this filter without using custom logic Click OK DPL refreshes the Endpoint Database restricting it further to endpoints in which Actual Price is Low or Nominal Custom filters can be used to drill down further in the Endpoint Database letting you see narrowly defined sets of endpoints you can combine AND with OR conditions and you can include more conditions than the five provided in the dialog 313 Chapter 11 Using the Endpoint Database Syncopation Software gt Click Grid Sort amp Filter Clear to clear the filter and display the entire Endpoint Database 11 1 2 Sorting the Endpoint Database You can sort the Endpoint Database via the Sort dialog or directly from the ribbon Within the Sort dialog you may sort using up to three columns Among many other things this feature is useful for finding out the maximum and minimum values of your objective function and which paths lead to these best case or worst case outcomes Select Grid Sort amp Filter
74. as shown in Figure 2 17 41 Chapter 2 Building Decision Analysis Models Syncopation Software Conditioning Revenue per Unit The values for this node depend or Product Strategy Number of value expressions required 2 Figure 2 17 Conditioning Dialog gt Click OK to close the Conditioning dialog The Node Definition dialog Data tab now indicates that two values are required for Revenue per Unit one for the In house alternative of the Product Strategy decision and a second for the License alternative DPL has copied the value that was specified when the Revenue per Unit node was not conditioned on anything to both entries DPL indicates the currently selected entry by coloring it magenta in the data input tree With the In house alternative selected click the select Variable button M to display the Select Variable dialog See Figure 2 18 42 Syncopation Software Chapter 2 Building Decision Analysis Models Select Variable Cost_per Unit Launch Costs License _Fee Marketing Costs NPY Price Product Strategy Unit_Sales From programs Figure 2 18 Select Variable Dialog Double click Price in the list DPL replaces the number 2 with Price within the data input tree Note In variable names DPL replaces spaces and punctuation marks with underscores _ in multiple word node names so the value node Unit Sales has a variable name of Unit_Sales The variable name is what is used in in e
75. decision alternatives with a policy dependent probability of zero are not displayed The information displayed in the policy tip is also available in the Session Log for all the points gt Make the Workspace Window wider by dragging its edge and the Session Log longer by dragging the splitter between the two panes The expected value of the objective function and the policy information is written to the Session Log for each point See Figure 14 9 382 Syncopation Software Chapter 14 Sensitivity Analyses Session Log Per_Unit_Costs 0 8 License_Fee 0 Expected value 35 1 EW Price_Change Raise 30 Keep same 70 does not occur 0 Product_ Strategy In_house 100 License 0 does not occur 0 Per Unit_Costs 0 8 License Fee 0 T Expected value 35 1 EV Price_ Change Raise 30 Keep same 70 does not occur 0 Product Strategy In_house 100 License 0 does not occur 0 Per_Unt_Costs 0 8 License_Fee 0 8 Expected value 35 1 EV Price_Change Raise 30 Keep same 70 does not occur 0 Product_ Strategy In_house 100 License 0 does not occur 0 Per_Unit_Costs 0 8 License Fee 0 8 Expected value 35 1 EW Price_ Change Raise 30 Keep same 70 does not occur 0 Product Strategy In_house 100 License 0 does not occur 0 Per_Unit_Costs 0 8 License_Fee 0 3 Expected value 35 1 EV Price_Change Raise 30 Keep same 70 does not occur 0 Product Strategy In_ house
76. default alternatives using the General tab of the Node Definition dialog 364 Syncopation Software Chapter 13 Risk Tolerance 13 RISK TOLERANCE While there are theoretical arguments for making decisions based on expected value alone often decisions are made with some consideration of the degree of risk involved in the different alternatives In many cases good risk visualization is enough and a judgment about risk can be made after looking at Risk Profiles Time Series Percentiles Tornado Diagrams and other results In situations where explicitly modeling attitudes toward risk is appropriate DPL gives you the ability to model your risk aversion or risk seeking in terms of a utility function To model attitudes toward risk using an exponential utility function a standard form well supported by research you need only specify a single number the risk tolerance coefficient This chapter assumes you have already been through at least one of the earlier tutorials Chapters 1 through 5 in this manual or that you are already familiar with DPL Therefore basic procedures such as starting DPL and opening and saving workspaces will not be explained in detail here 13 1 INCORPORATING A RISK TOLERANCE You ll start with a file containing a model similar in concept to the one introduced in Chapter2 Start DPL Open RiskTolerance da in the Examples folder underneath where DPL is installed C Program Files x86 Syncopation DP
77. events are allowed to vary across their states The base run that the results are compared to will equal the expected value of the model as in a Risk Profile or a Policy Tree Probabilistic Model A model in which some inputs are described with probability distributions and which generates probability distributions as outputs A probabilistic model in DPL has at least one chance event If the model has been developed graphically in the Model Window then a probabilistic model will have either continuous or discrete chance nodes or both 471 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software Program Window Provides an interface for creating editing and compiling DPL program files Converted spreadsheets can be viewed in this window Promoting Terms Moving elements of get pay expressions to nodes which are encountered earlier in the evaluation The advantage to this approach is that terms can be calculated in the tree as soon as all conditioning events are in known states rather than waiting until the end of the path to calculate all the terms at once In general promoting terms reduces run time R Rainbow Diagram An analysis tool which varies a single value over a user specified range and displays the impact on the expected value or certain equivalent and changes to the optimal decision policy Rayleigh Distribution A named distribution supported by DPL For a detailed description see On Line Help Re
78. expected value result from when you ran the model in Section 2 3 So the model has not changed much from how it was originally constructed The optimal decision alternative is to produce the product in house When the Policy Tree is created and activated a new ribbon tab labeled Policy with a yellow context shading and outline appears Policy Trees and the commands within the Policy tab will be covered in more detail in Section 3 1 52 Syncopation Software Chapter 2 Building Decision Analysis Models 2 5 VALUE TORNADO DIAGRAM You may want to know how sensitive the decision to produce the product in house is to changes in value for the variables in your model This can be accomplished by running a Value Tornado Note that since the model is deterministic has no chance nodes only value and decision nodes the Value Tornado diagram is the only type of tornado diagram you can run at this point gt Click Home Sensitivity Tornado Value Tornado is the default currently indicated by the icon or press F8 F8 is the shortcut to run the default tornado type You will see a blank Value Tornado Setup dialog similar to Figure 2 28 the figure shows the dialog with the values already entered The Value Tornado Setup dialog provides a table in which to enter the low and high settings range for each value node that is not a formula The middle column shows the current setting for each variable The Cost per Unit and Mark
79. followed by the expected value of each of the two attributes For example if the optimal policy Core Sample is followed and Seismic Structure is Open the expected value of the objective function at that point is 27 6 the expected values of the Profit and Land attributes are 33 6 and 12 0 respectively If you did not want to see the expected value of each attribute at every node you can turn off the display of Fat Policy on the ribbon by unchecking the Policy Display Show Attribute EVs checkbox 245 Chapter 8 Objective Functions and Multiple Attributes Syncopation Software You can also hold your mouse cursor over any node to see the expected values associated with that node more clearly in a tips box See Figure 8 11 Drilling Costs Drill Yes 27 6 33 6 12 0 27 6 33 6 12 0 cia i Chance Seismic Structure State Closed Level 3 EV of Objective Function 84 2 Drill Expected value of attributes 90 2 12 0 64 2 90 2 12 0 6 0 6 0 0 0 Figure 8 11 Policy Tree Displaying Fat Policy and Tips Finally you will run a Rainbow Diagram on the Land Weight value You want to know how much more penalty there would need to be associated with land disturbed for the optimal alternative to change i e how much more negative Land Weight would need to be Click Home Sensitivity Rainbow Diagram The Rainbow Diagram dialog appears Click Select The Select Value for Rainbow dialog appears Sele
80. for any cells or ranges that may later become calculation or initialization linked to value chance or decision nodes in DPL You may also find it helpful to adopt standard conventions for your range names although DPL does not require this When a DPL model is linked to a spreadsheet DPL uses Excel to recalculate the outputs e g NPV in each scenario This approach is easy and simple and works well with spreadsheets of moderate complexity and size Very large or complex spreadsheets may require a long time to recalculate Solutions for using DPL with spreadsheets of advanced complexity such as spreadsheet conversion are described in the Chapter 16 437 Appendix A Overview of Spreadsheet Linking Syncopation Software A 2 GALCULATION LINKS There are three ways to establish calculation links between a DPL model and an Excel spreadsheet A table summarizing the three methods is given at the end of this section A 2 1 Method 1 Creating a New Calculation linked Model from Excel When you are starting with a more or less complete Excel spreadsheet but have not yet developed a DPL model for the spreadsheet you may want to use the command Home Add to WS Excel Linked Model This command creates a DPL Influence Diagram from the named cells in an Excel spreadsheet all or a specified selection of them DPL creates nodes for the selected cells and influence arcs between export nodes and import nodes When you do this you wil
81. have already been through the tutorial contained in Chapters 2 and 3 or that you are already familiar with DPL To learn more about how to use a strategy table node consider the following example Select File Open Navigate to the Examples folder underneath where you installed DPL If you used the default location the path is C Program Files x86 Syncopation DPL8 Examples Select Strategy Table da DPL opens the Workspace as shown in Figure 4 1 125 Chapter 4 Strategy Tables Syncopation Software 9 d DPL Strategy Table da dev plan no table OEN Mode Mortle Catio Sampiet Rename ia CODY pf Find Replace n Objective Function Polery Samary Endpoints Decuion Workspace Manager sif Workipece net n we dev plan a table E dev punds g gt dev plan table 5 Ta t WE Jr R j i fo an ay a N x ta Se r a gt dev plan motable X On dev plin table X Ij dev plansis X ap Figure 4 1 Strategy Table Model The sample model is for a company with three product lines A B and C The company must make product development decisions called Feature Set A Feature Set B etc for each product line pricing decisions for each product line and a decision regarding how much to spend on marketing The company is faced with uncertainties in the demand for its products as well as uncertainties regarding how much it will cost to produce the updated product lines
82. help determine which variables have the greatest effect on the objective function and or are decision sensitive if applicable Those variables with the greatest impact or which are decision sensitive may subsequently be modeled as uncertainties When used with probabilistic models Value Tornado Diagrams can only be run on individual states of events or conditioned values For probabilistic models Value Tornado Diagrams may not be as useful as Event Tornado Diagrams or Base Case Tornado Diagrams 4 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software W Weibull Distribution A named distribution supported by DPL For a detailed description see On Line Help Workspace Manager A pane in DPL located within the Workspace Window which lists all documents and stored Risk Profile data associated with the current Workspace The Workspace Manager can be used to rename or delete items access saved data or switch to another window 4 8 Syncopation Software INDEX alternative outcome grouping ASVMMECWCviciiascciaccnnacaiwdsecwiennes 187 MIM e P E EE deers 210 ILEI E E E 253 408 asymmetric alternative outcome GlOUDING kriaepanrar ikad 187 aASyMMEeEtr C tree ssssesssssssrnrrersnnn 183 attachment points s sesesessnenennn 189 Attributes multiple See multiple attributes Base Case Tornado cccceeeeeeeees 67 Bayesian revision cscsceeeeeeeeeeaes 215 blue triangle c
83. in Chapters 2 and 3 gt Drop down the Home Sensitivity Rainbow split button and select Two Way The Run Two Way Rainbow Diagram dialog appears as shown in Figure 14 2 376 Syncopation Software Chapter 14 Sensitivity Analyses Two Way Rainbow Diagram License vs In house downstream First Variable axig Vanable name Current value Mone From Number of steps 5 Step size Second Yarable y aris Variable name Select Current value None From Number of steps 5 C Step size Evaluatian method Fast sequence evaluation OK Cancel Help Figure 14 2 Run Two Way Rainbow Diagram Dialog Click the Select button in the First Variable x axis section The Select Value dialog appears as shown in Figure 14 3 Value for sensitivity Current value Figure 14 3 Select Value Dialog for Two Way Rainbow gt Select Per_Unit_Costs in the Value for sensitivity drop down list 377 Chapter 14 Sensitivity Analyses Syncopation Software The Select Value dialog updates to show you that the value Per_Unit_Costs is conditioned as shown in Figure 14 4 In this instance Per Unit Costs is a discrete chance node that has probability and value data Therefore the value for Per_Unit_Costs is conditioned by which state or outcome the Per Unit Costs discrete chance node is in To run a Two Way Rainbow Diagram on a value that is conditioned you must specify the conditioning states fo
84. is 40 The expected values at the ends of these branches are 24 4 and 9 2 respectively So rolling back the tree to calculate the expected value where Product Strategy is In house and Unit Sales is low you take 60 24 4 40 9 2 18 3 Value is displayed in square brackets e g 107 7 Table 3 5 Node Information Displayed in the Policy Tree 82 Syncopation Software Chapter 3 Decision Analysis Results Data element location key Definition State name 3 Branch probability 4 The conditional probability of this The probability can either be discrete chance outcome occurring on this path If the displayed as a percent between discrete chance is conditioned b 0 100 or as a number y between 0 1 This data is other nodes this is the probability i of the chance outcome occurring displayed for Discrete Chance i h hi h nodes only given the outcomes ont S pat of the nodes that condition it Get pay values 5 The value of the get pay expression if any that occurs at this point in the tree If there is no get pay value at this point in the tree then nothing is displayed Table 3 6 Branch Information Displayed in the Policy Tree In the model you have developed so far there is only one get pay expression NPV in the tree on the branches of Price Price is the last discrete chance node in the tree furthest to the right In the Policy Tree the endpoints follow it Therefor
85. is causing the change in outcome of the other variable Lack of causation may be due to for example both variables being influenced by some other variable which is not in the model You will now re run the model from Section 3 4 to generate Value Correlations Make sure the License vs In house downstream model is active Check the checkbox next to Value Correlations in the Home Run group Leave the other outputs selected if you would like to re generate them gt Click the Home Run Decision Analysis split button Note when you request value correlations DPL must evaluate the model using full tree enumeration and does this automatically This changes the default action of the Home Run Decision Analysis button For this simple example it is unlikely you will notice the difference For larger models you may wish to return to fast sequence evaluation after running value correlations Use the Home Run Decision Analysis drop down list to do this the next time you run the model DPL runs the model and creates an item for Value Correlations in the Workspace Manager Double click on the Value Correlations item to activate the window The Value Correlations chart is displayed as in Figure 3 29 110 Syncopation Software Chapter 3 Decision Analysis Results Correlation Coefficient NPV Total Unit Sales Sales 1 Sales 2 Price Change Revenue per Unit Price Cost per Unit Marketing Costs E Positiv
86. is to convert an existing Influence Diagram Decision Tree model to a program and inspect the code that results As you become more proficient with DPL you may find it helpful to convert your models to programs from time to time for debugging documentation or other purposes In this section you will convert the energy case to a DPL program Select File Open gt Open the energy case file that you also used in Chapter 6 The default location path and filename is C Program Files x86 Syncopation DPL8 Examples Energy Case da gt Select Home Workspace Add to WS Program from Model DPL converts the model to a DPL program and displays the converted model in a Program Window Note that DPL did not replace the model with the program both the model and the program are contained in the Workspace and can be seen as separate items in the Workspace Manager Below is the text of the DPL program for the energy case The text is included here for reference but it will be easier for you to read in the DPL Program Window because of its width Use the scroll bar at the bottom of the window to scroll right and see all the text If you like select File Print and print the text of the energy case program You may need to print in Landscape format to see the entire program string Excel 1 C Program Files x86 Syncopation DPL8 Examples Energy Case xls excel Excel 1 Assumpctions Discount Rate value Discount Rate
87. it again You may wish to save the workspace first For the Decision Analysis options within the Home Run group make sure Risk Profile is unchecked gt If Policy Tree is unchecked check it The Home Run group should look like Figure 2 26 50 Syncopation Software Chapter 2 Building Decision Analysis Models C Risk Profile Policy Tree 2 all 7 4 none T C Policy Summary T Endpoints Decision Analysis Init Dec Alts T YOIC E Value Correlations Run Figure 2 26 Decision Analysis Options for Model with Decision Click Home Run Decision Analysis or press F10 F10 is the shortcut for the default run type indicated in the Home Run Decision Analysis icon DPL runs the model with the newly added decision node and produces the Policy Tree output that you requested See Figure 2 27 wv Sa lt y DPL New Product Strategy da Policy Treet Ea Workspace FA Fort ow _ we likeme ve in howe Product Strategy In house 105 1 105 1 sasalo tt License 93 7 937 A Figure 2 27 Policy Tree for Model with Decision 51 Chapter 2 Building Decision Analysis Models Syncopation Software DPL has created an item in the Workspace Manager for the Policy Tree which displays information in a decision tree format about the results of the decision analysis The first thing to note is the structure of the Policy Tree It displays the Product Strategy decision
88. items in the Workspace The Run group and Sensitivity group display the analyses that can be run on models and the options associated with those analyses Model Tab The Model tab contains a number commands for editing DPL models The buttons their names and the action associated with each are described in the tables below 12 Syncopation Software Chapter 1 Introduction Model The dialog box launcher will open Sum Node Options dialog where advanced Value and Decision node features can be turned on off S C TT Product Edit Change Tor A Difference Node Walue Add Node Discrete Chance Add a new node to the Influence Diagram of the default type indicated by Decision the icon Use the drop down split to select node type other than default The default changes to the last added type Continuous Chance n en J Strategy Table LL Controlled Collection Edit Node Edit the selected node s name states or dimensions data and or links Value Discrete Chance Continuous Chance Decision Controlled A Difference Table 1 2 Model Change to Node type Change type of selected node to the default type indicated by icon Use drop down split to select node type other than default The default changes to the last changed to type Sum Difference Product Create a new node that is the sum product or difference of the selected nodes Node Ribbon Group 13
89. just did and then look at the format and the comment lines at the top of the file Also if you intend to do any post processing of the exported Endpoint Database it may be easier to do so with one that has been exported with state numbers In the next section you will see how to import endpoints 11 5 IMPORTING ENDPOINTS There are two ways to import endpoints from a DPL Workspace DA file or from a CSV file In addition you can import and merge multiple Endpoint Databases This section contains three brief tutorials to demonstrate how importing works When you import an Endpoint Database DPL does not display to you any endpoint probabilities contained in the database These probabilities are not used when you play endpoints Further if the endpoints were generated in an application other than DPL probability data will most likely be missing 11 5 1 Importing from DPL Workspace Files You can import an Endpoint Database from a previously saved DPL Workspace DA file and connect it to your model As you might expect the Endpoint Database structure must match the structure of your model in order for DPL to be able to successfully import and play the imported endpoints In this tutorial you will continue with the energy case model but you will import an Endpoint Database containing different results You should still have the energy model open if not open Energy Case da from the folder where you installed
90. method of evaluating models based on Discrete Tree Simulation In Distributed Sampling the probabilities of nodes in the tree are represented exactly until the number of samples remaining at a node is small At this point normal Discrete Tree Simulation selecting chance node outcomes randomly is used for the rest of the tree Distribution There are two types of probability distributions in DPL 1 Chance node data specified as set of discrete states and associated probabilities 2 Chance node data specified as a named distribution See Named Distributions DLL See Dynamic Link Library Don t Gamble A property assigned to a chance node that effectively reduces it to one branch DPL replaces a chance event with its expected value if the Don t Gamble specification has been set Subsequent variables depending on the state of this event will be assigned values calculated by taking the expected value over the event where feasible Downstream Decision A decision which occurs in a decision tree after initial uncertainties are resolved Downstream decisions are sometimes called real options DPL Program A description of a decision model in DPL s own language i e DPL Code It contains all the structure and data necessary to analyze a model and can be used in place of an Influence Diagram and decision tree or as documentation for a graphic model 464 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms Dum
91. model The Policy Tree in Figure 3 2 also differs from the previous Policy Tree in Figure 2 27 because it now contains more levels DPL indicates this by displaying discrete chance nodes at the end of each decision alternative branch of the Product Strategy decision Recall in the previous Policy Tree that there were endpoints at the end of each of the decision alternative branches To look further at the Policy Tree you will expand the discrete chance node at the end of the In house branch 78 Syncopation Software Chapter 3 Decision Analysis Results Select the Cost per Unit discrete chance node at the end of the In house decision alternative branch The Expand split button within Policy Display group becomes active The Expand split drop down lists five menu commands for changing the state of expansion of the Policy Tree If you were to right click on the Same node in the Policy Tree to bring up the context menu you will find four of the same expand commands with the exception of the Expand All command All are summarized in Table 3 4 below Control Action Expand Expands the currently selected node one level if the node has never been expanded i mi before or to the previously expanded state This can also be accomplished by double clicking the node Expand and Zoom Does the same thing as Expand but then zooms the Policy Tree so that the whole A tree can be seen This can also be pe
92. node data Your model should now have all the arcs shown in Figure 2 10 33 Chapter 2 Building Decision Analysis Models Syncopation Software Figure 2 10 Model with All Influence Arcs Note the above process provided you with some familiarity on how to create individual Excel linked nodes and how to create influence arcs This information is useful when building and modifying DPL models However DPL also provides a command which will create the model shown in Figure 2 10 in a single step Before continuing you will try this now Select Home Workspace Add to WS Excel Linked Model In the Create Model from Excel dialog click Browse and select the file New Product Strategy xls It will be in the folder with DPL examples usually C Program Files x86 Syncopation DPL8 Examples Select the same cells as shown Figure 2 3 Click OK ov 34 Syncopation Software Chapter 2 Building Decision Analysis Models Note DPL has created a new model in the Workspace called New Product Strategy i e the same name as the spreadsheet Other than node arrangement this model is identical to the model in Figure 2 10 This is a much quicker way to build a model from an Excel spreadsheet However it is also important to know the steps outlined in Sections 2 1 and 2 2 Select the New Product Strategy model in the Workspace Manager Press Delete gt Click Delete Model in the Confirm Delete dialog You will now continue w
93. not connected to a model the report and the Endpoint Database it displays will be deleted I e it is not possible to have a disconnected Endpoint Database that is not displayed in a report The original Endpoint Database still named Endpoint Database is currently connected to your model You will now change your value model and re run over writing this database gt Switch to Excel to edit the Energy Case xls spreadsheet gt In the Assumptions tab change both the Annual Price Growth assumptions cells C24 and C25 to 5 Also change the Annual Op Costs Growth assumptions cells C26 and C27 so that they are both 2 gt Change the value of Initial Marketing Spend cell C33 to 650 The Other Assumptions section of the spreadsheet should look like Figure 11 19 Other Assumptions Annual Price Growth Market 1 Annual Price Growth Market 2 Annual Op Costs Growth Market 1 Annual Op Costs Growth Market 2 Start year 2013 value p value q Upfront Cost Escalation Initial Marketing Spend Upfront NPV Total NPV 41 M 4 gt gt Assumptions lt DCF lt Market Trend Data 3 4 Ma Ready 7 A GI 110 C g Figure 11 19 New Assumptions in Energy Case xls 328 Syncopation Software Chapter 11 Using the Endpoint Database Switch back to DPL gt Make sure that Endpoints Policy Tree and Policy Summary are checked and run a Full Tree Enumeratio
94. objective function value but as License Fee increases from 0 85 to 0 9 there is an increase in the objective function value 114 Syncopation Software Chapter 3 Decision Analysis Results m w Ha _ v e gt U D a U D Q x WwW 0 8 License Fee Figure 3 33 Rainbow Diagram for License Fee Rainbow Diagrams also indicate policy changes by changes in color Since the Rainbow Diagram in Figure 3 33 changes color between 0 85 and 0 875 you know that a policy change occurs in that region As with Tornado diagrams the vertical line where the color changes does not indicate the precise point at which the policy changes The vertical line is drawn at the halfway point of the two tested values of the variable between which a policy change has occurred In this instance License Fee is tested at 0 85 which results in one policy and again at 0 875 which results in a second policy Therefore the vertical line is drawn halfway between 0 85 and 0 875 If you want more precise information on where the policy change occurs you can run another Rainbow Diagram with finer increments In a Rainbow Diagram DPL starts with a color and changes to a new color every time there is a policy change The color of the region does not relate to the value of the variable i e blue does not mean that the high setting of the variable resulted in a policy change It only indicates that a policy change has occurred You can
95. of the sensitivity variables are tested i e with the model as currently defined in the Model Window The Base Result also establishes the optimal policy for the model Bayes Rule A mathematical method for reversing the order of conditioning in chance events e g used to assign probability when you have the probability of A given B but need the probability of B given A The formula for Bayes Rule is P Ai B P Ai xP B Ai P A1 xP B A1 P A2 xP B A2 P An xP B An Beta Distribution A named distribution supported by DPL For a detailed description see On Line Help Binomial Distribution A named distribution supported by DPL For a detailed description see On Line Help Branch An element in a decision tree representing one state of an event Decision node branches represent alternative choices and chance node branches represent possible outcomes Branch Block A modeling option that temporarily removes one or more alternatives from consideration in a given decision node instance Branch Control A modeling option that forces the outcome or branch of an event to a particular state in a decision tree 460 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms C Certain Equivalent CE The certain amount equivalent to the expected utility of the model taking risk attitude into account Specifically the minimum guaranteed amount a decision maker would accept in place of an uncertain lott
96. on Cost per Unit Marketing Costs Price EET Strategy 5Sales 1 Number of value expressions required 6 Cancel Figure 3 17 Conditioning dialog for Sales 2 value Note that the Conditioning dialog tells you how many data expressions are required given the nodes selected Since Sales 2 is a three outcome discrete chance node and Price Change is a two alternative decision node six data expressions are required 96 Syncopation Software Chapter 3 Decision Analysis Results gt Click OK to close the Conditioning dialog The data input tree updates on the Values tab to account for the conditioning Node Definition Sales 2 General Conditioning Print Page Setup Copy FullScreen whee Reorder QQA L visi Probabilities Links Low Sales 8 Raise 2 Nominal 15 Price High Change 28 Low Sales 10 Keep same 2 Nominal 20 High Figure 3 18 Node Definition Value Tab After Conditioning gt Change the Low Nominal and High values for the Raise alternative of Price Change to be 8 15 and 28 respectively You are assuming that demand sales is slightly sensitive to price Leave the values for the Keep same alternative as is The data input tree for Values should now look like Figure 3 18 Select the Probabilities tab Click Conditioning The Conditioning dialog appears Click Sales 1 as shown in Figure 3 19 Click OK LUUL 97 Chapter 3 Decision Analysi
97. on events decision Strategy US Europe string Currency given Strategy Dollars Euros You can use a string name in place of a string literal anywhere that DPL allows string literals You can use strings and string literals as the argument of a display function as arguments of the lookup functions hlookup vlookup and as the arguments of the relational operators If strings are used as arguments of the relational operators lt gt and so on the ordering is determined by the standard ASCII values for characters For example Both relations are true that is they both result in a value of 1 If a string name appears in a math expression DPL will generate a Not a number error message 15 4 12 Named Constants A named numeric constant represents a number When you define a named constant you assign it an identifier and an initial value DPL stores the initial value at the time it compiles a program and it does not allow the value to change thereafter The following statement defines a named numeric constant in DPL const BaseYear 2014 410 Syncopation Software Chapter 15 DPL Programs and Code You can use a named numeric constant or constant expressions containing named numeric constants in places where a variable is not allowed These include DPL statements that define the number of attributes associated with alternatives and outcomes the number of alternatives for a decision the number of ou
98. once you have specified Excel as the Calculation link type you can copy the cell in Excel using the Excel command Home Clipboard Copy or Ctrl C switch back to DPL and click the Paste Link button on the Links tab of Node Definition dialog See Figure A 4 and Figure A 5 442 Syncopation Software Selecting Calculation Link Hame LOGS Develop Discount_rate FOA_ approval Key competitor outcome launch year Market_share Market size Marketing cost rate patent expiry Phase 1_ payment Phase 1_ success Phase 2 payment Phase 2 success Phase 3 payment Phase 3 success Pricing prob fda_approval prob success phased prob success phases Registration payment royalty pabe Start ear Years _to_peak Appendix A Overview of Spreadsheet Linking Cell Names Asymmetric rees xls Location Assumptions 0945 Assumptions 0919 Assumptions C 6 Assumptions o 24 Assumptions 0926 Assumptions 049 Assumptions C Assumptions 0926 Assumptions o 42 Assumptions C Assumptions l 51 Assumpltionsl l 21 Assumptions C 52 Assumptions o 22 Assumptions 0955 Assumptions l 23 Assumptions 0929 Probabilities 0912 Probabilities tC E Probabilities 411 Assumptions C 54 Assumptions 0550 Assumptions 098 Assumptions 0959 Count l l l l l l b l l l b l l l l l l l l l l l l l Go to Cell Cancel Sort by location Show cells already linked
99. over the green circle for 0 04 as has been done in Figure 5 35 DPL will tell you that Spin off is Yes in that step For more information on reading policy tips please see Section 3 7 Note Show Tips in File Options General must be checked in order for policy tips to work 178 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models 5 10 MODELING A DOWNSTREAM DECISION The decision analysis model used in the previous two sections assumes that the Spin off decision will be made at a time when the four uncertainties are not known In Figure 5 30 the decision comes first in the decision tree sequence You might ask how the results would differ if the decision could be made later after the uncertainty Share of at Risk Revenue Lost is resolved To answer this question you will change the sequence of the nodes so that the decision is preceded by this uncertainty A decision that comes after one or more uncertainties is called a downstream decision or a real option For more on modeling real options using downstream decisions see Chapter 7 gt In the Workspace Manager right click on Decision Analysis Select Duplicate Right click on the new model Select Rename and call it Real Option Analysis Make the new model active in the Model Window To make the Spin off decision a real option triggered by the Share of at Risk Revenue Lost uncertainty you will need to tell DPL that the uncertai
100. pane at the top and 2 the Session Log pane at the bottom The Workspace Manager pane displays all the documents in the open Workspace in a Windows Explorer style tree structure The top level is the Workspace Below the Workspace will be displayed all the models in the Workspace Below each model will be displayed the outputs and other items associated with each model Currently there is only one model Model1 in the Workspace Right click on the icon for Model1 to see the context menu for a model Notice that you can delete rename or duplicate a model Syncopation Software Chapter 1 Introduction Within a DPL Workspace you can develop run and store results and outputs from multiple models The model you most recently ran compiled or viewed is bolded in the Workspace Manager In general when you run a Decision Analysis or use any other run or compile command DPL runs or compiles the active model Once you ve run or compiled a model its item in the Workspace Manager will have a compiled indicator double chevrons gt around the model name The Session Log displays information about the current session Presently it displays information about the version of DPL you are running and to whom the software is licensed See Figure 1 2 You can also access DPL license information by clicking Help Resources License As you build models and run analyses further information such as compilation messages and the expected
101. pitfalls to be aware of Chapters 2 and 3 discuss the basics of DPL linked spreadsheets how to build a model from Excel and how to modify it by adding linked nodes to the model Throughout those chapters the model uses Excel as its calculation engine for the cash flows that DPL rolls up in the decision tree DPL communicates with Excel either via OLE Automation or DDE Messaging depending upon the communication protocol selected in the Settings tab of the Model Links Options dialog If you have a large model these communications protocols can be slow Also DPL s analytical engine is optimized for performance on decision trees with numerous paths whereas Excel is designed for a single interactive recalculation For these reasons if you have a large model and wish to substantially reduce runtime you may want to convert your spreadsheet to code 16 1 HOw TO CONVERT SPREADSHEETS Select File Open Navigate to the Examples folder underneath where you installed DPL If the default location was used for installing the path is C Program Files x86 Syncopation DPL8 Examples Select Software da DPL opens the Workspace as shown in Figure 16 1 425 Chapter 16 Converting Spreadsheets to DPL Code Syncopation Software Sn Gio DPL Software cash flow decison gt mcii File Tomado Time Series Weorkipace Manager g F Softwareds _ tash flow decision Licenses User suser Sync
102. please contact technical support 15 6 1 What is a Compiler Directive DPL code is not procedural This means that a DPL program simply describes a decision problem but does not include instructions on how to solve the problem In this way DPL is more like a spreadsheet than C or Java which are procedural However although a DPL program does not contain instructions for solving a decision analysis problem it can contain directions to the compiler that control the process by which the compiler reads a DPL program and prepares it for analysis by the DPL engine 421 Chapter 15 DPL Programs and Code Syncopation Software These instructions are called compiler directives Every compiler directive begins with a symbol in the first column of a line of a program file There are compiler directives that e Include additional programs in a compiled program e Leave parts of a program file out of the compiled version e Control optimization and spreadsheet evaluation 15 6 2 Including the Contents of One DPL Program in a Second Program DPL code includes a compiler directive that allows a DPL program to be broken into several parts This can allow you to maintain a list of constants that you use in several models or use a single value model in several programs or update a spreadsheet model convert it to code again and run a program without editing the DPL models or program The compiler directive to include a program is includ
103. program If you generate a DPL program from a graphical DPL model the Influence Diagram will be described in a definition section The Decision Tree will be described in a sequence section Most programs created from graphical models will have both a definition section and a sequence section If you convert a spreadsheet to a DPL program file it will have the appropriate format to be included in the definition section and it will not contain a sequence section 392 Syncopation Software Chapter 15 DPL Programs and Code 15 2 1 The Definition Section Most DPL programs start with a definition section This section which specifies and initializes the components of a decision problem usually contains the bulk of the model There are two components you can define in the definition section e Events which include decisions chance events and controlled events e Values which include constants scalars series arrays with string or numeric data Strategy tables are handled as a set of decisions in DPL programs 15 2 2 The Sequence Section DPL sequence sections are systematic English language descriptions of Decision Tree structures There are two tasks to accomplish in the sequence section of a DPL program First you must specify the chronological order in which the events or nodes occur Second you must specify the actions such as payment or receipt of values which are associated with each branch 15 2 3 Comments Com
104. ribbon or via the Control tab of the Branch Definition dialog To control via the ribbon select the branches of the Actual Reserves node in the decision tree and select Low from the Control drop down list in Model Branch See Figure 11 16 Ctrl state Control All Uncontrol All Branch Figure 11 16 Branch Control via the Command Ribbon To control branches via the Control tab of the Branch Definition dialog double click the branches of a node in the decision tree and click the Control tab Select the Control radio button and choose the control state from the drop down list Also branch control the Market 2 Political Stability Future node to Unstable Your Decision Tree should look like Figure 11 17 325 Chapter 11 Using the Endpoint Database Syncopation Software Actual Actual Operating Pipeline Price Reserves Costs S Market 2 Appraisal Forecast Paica Mkt 1 10 BCM Total_NPV b nits Results Price r men Nominal Nominal Nominal Nominal Investment aad Low Low is E Mkt 1 20 BCM Total T Low Stable Ya High High High High AR Nominal Nominal mm Market 2 Total_NPV High Unstable Political High Stability High Stability Future Mkt 1 and 2 10 BCM Stable Mkt 1 and 2 20 BCM Moderate Mkt 1 and 2 30 BCM Unstable Upfront_NPV mw Figure 11 17 Decision Tree with Branches Controlled for Two Chance Nodes gt Uncheck Risk Profile gt Make sure Policy Tree and Policy Summary
105. risk profile of the model to help you select which range you would like to specify In the Min edit box for Filter range type 200 Leave the Max edit box for Filter range at its specified value Note that the low and high for the range are inclusive values DPL will filter based on the objective function being greater than or equal to the low value and less than or equal to the high value Click OK By leaving the high setting for the filter range at the high setting for the full range you are asking DPL to filter the Endpoint Database for all optimal endpoints with an objective function value greater than or equal to 200 DPL creates another filtered Policy Summary as shown in Figure 12 17 359 Chapter 12 Advanced Decision Analysis Results Syncopation Software Test Drill Amount of Oil 0 Yes 100 gt 100 gt Core Sample Exp Seismic lt 0 gt 0 does not occur does not occur does not occur 55 gt Drilling Costs Oil Price Pipeline Costs Low lt 2 15 gt Low 0 50 gt Low 30 gt 74 gt 34 56 gt be 0 gt Nominal 60 56 gt 60 10 10 does not occur lt 0 gt Nominal m does not occur does not occur Seismic Structure Exp Seismic Test No 0 gt No 90 gt Closed lt 56 gt 24 lt 0 gt Closed does not occur does not occur Figur
106. s calculations e g unit sales or calculated results from the spreadsheet e g NPV D None local DPL Program Microsoft Excel Workbook c program files amp 86 syncopation dpl8 examples enlightify 1 x s Sheet Cell Assumptions Gross_Margin Go to Cell Cell Names y This is an export node Its value will be sent to Excel each time it changes during the run Initialization links These links are for connecting to a spreadsheet or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at the start of each run None local or DPL program Microsoft Excel Figure 5 9 Node Definition Links for Gross Margin gt Click the Go to Cell button This takes you to Excel and selects the linked cell Gross Margin in this case Switch back to DPL gt Click Cancel to close the dialog 149 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software 5 3 ARRANGING THE INFLUENCE DIAGRAM Before you begin to analyze the model take a moment to organize the Influence Diagram GUL 4 U Double click on the splitter bar between the two panes of the Model window This maximizes the Influence Diagram pane Click View Layout Snap to Grid Drag the Total Value node to the right Click in the Influence Diagram to deselect it While holding down the Ctrl
107. set at Yes The dialog should look like Figure 12 19 Option Value Chart Setup wildcat Decision Default Alternative Evaluation method Fast sequence evaluation Save above settings as Default Alternatives Use Defaults defined in Node General don t show dialog again Can be reset in Home Run Options OK cane Figure 12 19 Option Value Chart Setup with Changes Click OK DPL creates the Option Value Chart as in Figure 12 20 362 Syncopation Software Chapter 12 Advanced Decision Analysis Results Test Drill Figure 12 20 Option Value Chart for Wildcat Model DPL calculated the option values in this chart by first running the model with the Test decision controlled to Core Sample and the Drill decision controlled to Yes If this is always the strategy then the objective function value is approximately 11 which is the value at the bottom of the first bar on the chart DPL then ran the model with the first decision Test uncontrolled and the Drill decision still controlled to Yes If the Drill decision is always Yes then there is no value associated with testing so the optimal alternative for Test is None The objective function in this case is 17 which is the value at the top of the first bar and bottom of the second bar on the chart 363 Chapter 12 Advanced Decision Analysis Results Syncopation Software Finally DPL uncontrolled both decisions and ran the model again In this
108. states For Condition 2 select Actual Reserves Col_8 and select the Low outcome gt Leave the radio button for all selected The dialog should look like Figure 11 5 Filter Condition 1 Condition 2 Condition 3 Pipeline_Strategy Col_3 M Actual Reserves Col_8 none Mkt i and2 30 BCM Row must meet all 0 any of the above conditions Custom gt gt gt Figure 11 5 Completed Filter Dialog Click OK DPL displays the filtered Endpoint Database The endpoints shown in the Endpoint Database are filtered to the endpoints for which the Pipeline_Strategy is in either the Mkt_1__10_BCM or Mkt_1_and_2__ 10 _BCM alternatives and for which the outcome of Actual_Reserves is Low Using the filter dialog you can combine up to five filter conditions for the purposes of this manual only four are displayed above In the example above you defined a filter in which the endpoints must meet all of the conditions e g a logical AND between the conditions You may also define a filter in which the endpoints meet any of the conditions using the radio button below the conditions i e a logical OR is used between conditions in the filter You can also create a Custom filter using the Custom button as shown in the filter dialog Custom filters allow you to define more complex filtering rules so that you can view more specific subsets of the Endpoint Database To see how to do this you will modify the fi
109. that the variable that has the biggest impact on the objective function is at the top 55 Chapter 2 Building Decision Analysis Models Syncopation Software The changes in color indicate that a policy change has occurred for either the low or high setting of the variable In this model a policy change occurs when a different alternative is optimal for the upfront decision the decision at the head i e far left of the decision tree from that which is optimal for the Base Result policy We will further discuss policy changes in Sections 2 7 and 3 7 DPL indicates that a policy change has occurred for the low setting of the variable by displaying a portion of the bar for that variable in blue For a high setting policy change the end of the bar is displayed in light magenta The legend on the tornado diagram helps to explain this For different values the blue and or light magenta may occur on either side of the vertical line representing the Base Result In Figure 2 29 light magenta and blue both occur to the left of the Base Result This is because a low setting of Unit Sales decreases the objective function NPV while a high setting of Marketing Costs also decreases the objective function Note that DPL displays the change in color from green to blue or light magenta midway between the value of the objective function for the current setting and the value for the low or high setting For each variable DPL has only tested a low
110. the el1se and searches for a endif statement If the i expression is false equals 0 then the compiler searches for the first occurrence of a else statement or a endif statement It skips the code between the if statement and the else or endif statement If it finds a else statement first it compiles the code between else and the next endif Once the compiler finds a endif statement it starts compiling again 15 6 4 Setting Compile Time Options from Within a Program DPL allows you to control the settings of compile time options from a DPL program file Although these options can also be set using Model Options this feature can be very useful if forgetting to set an option results in long run times or unexpected results Like all compiler directives the options option list directive must begin in the first column of a line In general this directive should be used only once and near the beginning of a program file The option list consists of a sequence of letters separated by commas indicating the options to be set There are five compile time options that can be set in an options directive e e for expression optimization e s for sequence optimization e 1 for lottery optimization e x Excel lookup function convention e i ignore case when comparing strings If the letter is preceded by a minus sign the option will be turned off If the letter is preceded by a plus sign or nothing the option will be tu
111. the results of your Monte Carlo simulation including the minimum and maximum values the variance and standard deviation the higher moments skewness and kurtosis and selected percentiles of the distribution You may enter other percentiles into the Percentile list box and then click Update Percentile Values 163 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Statistics Distribution Expected Value kd Distribution statistics have also been Mean a written to the log Median 892 Percentile Val 1 2 Min ot 10 61 Max a0 892 30 1715 Vanance 99 140 none Standard Deviation none Skewness Kurtosis Figure 5 21 Risk Profile Statistics Click Close to close the Statistics dialog Note that because each simulation run will produce somewhat different results and you have run only 100 samples your results will not match Figure 5 21 exactly but they should be similar The Risk Profile statistics are also written to the Session Log To see this scroll down to the bottom of the Session Log 5 7 COMPARING RISK PROFILES DPL allows you to show up to eight risk profiles in a single chart window You ll use this capability to show how increasing the number of samples increases the accuracy of the results gt In the Workspace Manager right click on the Risk Profile Dataset item called Expected Value _ gt Choose Rename gt Call this
112. the Branch Definition dialog gt Click in the get or pay expressions cell for Attribute2 Click the Select Variable button dah and select Invest_1x5 from the list Click OK Click the Fill Down tah button in the Get Pay Definition dialog DPL will fill in the array elements Invest_1x5 0 Invest_1x5 1 and so forth for the Get Pay expressions The Get Pay Definition dialog would now look like Figure 10 23 If there is data in any of the subsequent cells you will get an overwrite data prompt before DPL fills in the subsequent values 302 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics Branch Definition Get Pay Proceed Get or pay expressions Control Attribute Get pay expression NPV Total _NPV Display Eee EEA 8 ee EAE ie i RS ee eee Attributed Invest_1x5 0 Attributes Invest_1x5 1 Attribute Inwest_1x5 2 Attributes Invest_1x5 3 Attribute Invest_1x5 4 Go to Attributes Get Pay Ll Wl j Cancel Help Figure 10 23 Get Pay Definition Dialog with Array Values Filled In The procedures for filling in Get Pay expressions with array elements using the Fill Down button are very similar for arrays with other dimensionality For example Suppose you have defined a 5x1 column vector array called Invest_5x1 Just as with the row vector you would select Invest_5x1 from the variables list and enter it for Attribute2 The Fill Down button tah will then fill i
113. the DPL examples 334 Syncopation Software Chapter 11 Using the Endpoint Database gt Switch to the Energy Case model You should have no Endpoint Database gt If needed press F9 to clear model memory and click OK for the warning With the model still activated select Home Workspace Add to WS External Endpoints Browse to find the file Energy Case for Import da in the path where you installed DPL This DPL file contains an Endpoint Database that was previously created with the energy model Select the file and click Open You will see the Import Endpoints dialog see Figure 11 25 Import Endpoints Browse Update Contents Merge with existing endpoints Display report Cancel Figure 11 25 Importing Endpoints from a DA File Leave Display report unchecked Click OK DPL imports the Endpoint Database You can see in the Workspace Manager that Imported Endpoint Database is present although the report is not shown gt Check Policy Tree and Policy Summary in the Home Run group 335 Chapter 11 Using the Endpoint Database Syncopation Software Run a Full Tree Enumeration from Endpoints The expected value of this model is approximately 50 and the policy is different from the energy model you have been using Note that you played endpoints instead of running the model If you were to run the model at this point you would revert to your earli
114. the Node Definition dialog with the Tree Instance tab active Double clicking in the branches of a node brings up the Get Pay tab of the Branch Definition dialog more on this later gt Double click on the Develop node in the decision tree not in the Influence Diagram to open the Tree Instance tab of the Node Definition dialog Select Asymmetric for the Alternative Grouping Figure 6 6 187 Chapter 6 Asymmetric Trees Syncopation Software General Display options Data V Show name Underline name C Border around name Links Altemative grouping Symmetric ga Asymmetric Figure 6 6 Node Definition dialog Tree Instance tab gt Click OK 188 Syncopation Software Chapter 6 Asymmetric Trees Pricing Low Low NPV Nominal Nominal Nominal High High Figure 6 7 Decision Tree with Asymmetric Decision DPL detaches the node from the rest of the tree and changes the decision node instance so that each alternative branch of the decision can have a different subtree follow it In Figure 6 7 the two blue triangles on the branches of the decision represent attachment points where other nodes can be connected You want to attach the FDA Approval node to the Yes branch Select FDA Approval by clicking it with the mouse Without releasing the mouse button drag the node until the semi transparent chance node is over the blue triangle for the Yes alternative Figur
115. the Ops Costs Research chance node in the Decision Tree gt Drag the node so that the chance node is positioned over the Proceed decision node and release the mouse button The newly added node should be positioned before the Proceed decision as shown in Figure 11 13 Actual Actual Pipeline Price Reserves Market 2 ees a Appraisal Forecast Sarre Ops Costs Mkt 1 10 BCM Total_NPV initia Results Pri Stability Nominal Nominal Nominal Nominal ESU Price Research Sk Proceed Mkt 1 20 BCM Investment Low n z Total A Stable vas High High High High Nominal Nominal Total_NPV Market 2 i Unstable High Political High Stability Future Mkt 1 and 2 10 BCM Stable Mkt 1 and 2 20 BCM Moderate Mkt 1 and 2 30 BCM Unstable Upfront_NPV Figure 11 13 Decision Tree with Reordered Op Costs Research Node You are ready to play endpoints with the conditioning node gt Drop down the Home Run Decision Analysis split and select Full Tree Enumeration from Endpoints or press Alt F10 321 Chapter 11 Using the Endpoint Database Syncopation Software You will see a warning message DPL is telling you that you have added a new event since the recording of the Endpoint Database If the conditioning relationship between the new event and existing events were not purely a probabilistic one you would not be able to play endpoints However the Op Costs Research node serves only one purpose which is to update the pr
116. the sensitivity event i e it is never replaced with its expected value in the Event Tornado Note because a Deterministic Event Tornado replaces chance events with their expected value it may not be appropriate to use for models that use a risk tolerance and hence have certain equivalents For such models consider using a Probabilistic Event Tornado You will now run a Deterministic Event Tornado If Product da is not already open select File Open Workspace Navigate to the Examples folder underneath where you installed DPL If the default location was used for installing the path is C Program Files x86 Syncopation DPL8 Examples Select Product da cy gt Inthe Home Sensitivity group drop down the Tornado split button and select Event Deterministic from the list DPL produces the Deterministic Event Tornado shown in Figure 14 10 385 Chapter 14 Sensitivity Analyses Syncopation Software Deterministic Event Tornado Sales Sales 2 Marketing Costs Per Unit Costs 15 20 25 30 35 40 45 50 55 Expected Value E 10th to 50th Percentile HI 50th to 90th Percentile Figure 14 10 Deterministic Event Tornado The width of each bar indicates the difference between the 10th percentile and the 90th percentile of the risk profile of the objective function for the model when the sensitivity event labeling the bar is the only uncertainty in the model unless chance events have been set
117. to Always Gamble The change in color occurs at the 50th percentile The bar is yellow between the 10th and the 50th percentiles and red between the 50th and 90th percentiles The black vertical line running through all the bars indicates the expected value established in the run for the Minimum Model Range A bar may not change color if the risk profile is highly skewed for a particular chance event For example if the 50th percentile were equal to the 90th percentile then the entire bar would be yellow The narrower a bar is for a chance event in a Deterministic Event Tornado the less impact the chance event has on the overall uncertainty in a model Bars nearer the top of a Deterministic Event Tornado contribute more to the overall uncertainty in a model 386 Syncopation Software Chapter 14 Sensitivity Analyses If a chance event appears in a Deterministic Event Tornado with a bar of no width then the 10th and 90th percentiles are equal The most likely reason for this is that Don t Gamble has been set for the event though it could be because the 10th and 90th percentiles are truly equal The latter may happen in an asymmetric tree when an uncertainty only occurs for a particular decision alternative and that alternative is never optimal You will now run a Probabilistic Event Tornado Inthe Home Sensitivity group drop down the Tornado split button again This time select Event Probabilistic from the list DPL p
118. to License out production of the product and to keep price the same in time period 2 To keep the policy tip compact decision alternatives with a policy dependent probability of zero are not displayed 116 Syncopation Software Chapter 3 Decision Analysis Results If you wish place your mouse cursor over one of the green circular markers on the curve to see what the policy is for these points You have completed a Rainbow Diagram on one value in your model namely License Fee You will now see how to run a Rainbow Diagram on a probability 3 90 RAINBOW DIAGRAM ON A PROBABILITY When you ran the Rainbow Diagram in Section 3 7 you may have noticed that the only variables you could select in the Select Value for Rainbow dialog were values What if you would like to run a sensitivity analysis on a probability You will need to modify your model slightly to do so LUUL Q Double click on the item for the License vs In house downstream model in the Workspace Manager The Model Window becomes active Add a value node to the model by clicking Model Node Add If the default add node type isn t value as indicated by the icon use the drop down list to select Value Click anywhere in the Influence Diagram The Node Definition dialog comes up with the General tab selected Name the value p Click the Data tab Type 0 6 as shown in Figure 3 35 117 Chapter 3 Decision Analysis Results Syncopation Software
119. tolerance coefficients for different paths through the tree You can define your own utility function for converting values into utiles If you do so you must also define the inverse utility function for converting utiles to values The expressions for the utility function and inverse utility function use a dollar sign to represent the arguments of these functions For example the following defines a utility function that is equivalent to the built in exponential utility function use utility exp RiskTolerance RiskTolerance log The expressions for the utility function and inverse utility function cannot contain values that depend on events Also these expressions cannot contain the state function The utility function defined above is the same as the default exponential utility function provided by DPL 418 Syncopation Software Chapter 15 DPL Programs and Code You are responsible for ensuring that the two functions are the inverse of each other DPL will perform an analysis with functions that are not the inverse of each other but the results may be meaningless 15 5 10 Multiple Attributes Objective Functions and Constraints You specify the number of attributes and an objective function if any immediately following the keyword sequence at the beginning of the sequence section of a program sequence attributes 2 objective 1 Cost 2 Cost2 The constant expression that represents the number
120. tree This position is normally used for converted spreadsheets Values defined in this program can be used as calculation links or in the decision tree Choose Choose ee remove Cancel Help Figure 16 2 Model Links Dialog In the Linked Spreadsheets section the Model Links dialog displays the list of Excel spreadsheets linked to the model In most instances there will only be one Excel spreadsheet linked to a model However you can have multiple spreadsheets linked to a model To establish a link to another spreadsheet you can click Create Link to New Spreadsheet In this instance there is only one spreadsheet link Note if no path is specified for the spreadsheet name then DPL assumes the spreadsheet is in the same directory as the Workspace file gt Make sure Cash Flow1 x ls is selected in the Linked Spreadsheets list box Click Convert to Calc Program The Convert Spreadsheet to Program dialog appears as shown in Figure 16 3 42 Chapter 16 Converting Spreadsheets to DPL Code Syncopation Software Convert Spreadsheet to Program Spreadsheet name Browse Options Cancel Figure 16 3 Convert Spreadsheet to Program Dialog The dialog has the spreadsheet name that you are about to convert in the Spreadsheet name box If needed you could change the spreadsheet by clicking the Browse button gt Click Options The Conversion Options dialog appears as shown in Figure 16 4
121. using DPL with Excel Select the cells named Cost_per_Unit Launch_Costs Marketing_Costs NPV Revenue_per_Unit and Unit_Sales as shown below Hint to select multiple names hold down the Ctrl key as you click the names 25 Chapter 2 Building Decision Analysis Models Syncopation Software Range Names C Program Files x86 Syncopation DPL Examples New Product Strategy xls Creating Calculation Linked Values Name Location Type Count Rows Col Base Date Oriversl C 12 Value 1 1 1 Select All Cost per Unit Drivers O 26 Value Discount Rate Drivers 0910 Value Launch Costs Drivers 0 30 Value Marketing Costs Drivers 28 Value Drivers 091 Formula Calculations C 24 Formula Revenye_per_Unit Drivers 22 Value Unit Sales Drivers 24 Value Cancel Sort by location ee l l l l l ee l l l l l l l ee ee l l l l l l l Show ranges already linked Show ranges with formulas Create arrays based on range SIZ Prefis node name with sheet name Figure 2 3 Cell Names Dialog Click OK DPL creates the new value nodes and adds them to the Influence Diagram Arrange the nodes so they look like Figure 2 4 below Drag the splitter down if you need more room 26 Syncopation Software Chapter 2 Building Decision Analysis Models Figure 2 4 Model with Six Value Nodes When you build a spreadsheet linked DPL model you will typically want to have DPL send dif
122. value of a model are written in the Session Log Error messages are also written in the Session Log The Session Log is cleared each time you close DPL and each time you open a new Workspace Session Log n DPL Decision Programming Language Copyright 2003 2013 by Syncopation Software Incorporating software on license to Syncopation Software Inc All rights reserved Www syncopation com support syncopation com US toll free 1 866 796 2375 Outside the US 1 978 233 2509 Professional Version Perpetual Release 6 00 00 Serial Number SY DP80130102001 Licensed User auser Syncopation Software Inc O2M4M3 11 23 22 Figure 1 2 The Session Log Chapter 1 Introduction Syncopation Software Both the Workspace Manager and the Session Log can be undocked by clicking the top of the pane and dragging away from the side of the window or by right clicking within the desired pane and selecting Dock The pane can be left floating or moved to one of the edges of the application window and re docked there The window border changes as you drag it to indicate whether and where it will be docked If you dock one pane on top of the other the window will show only the top pane and create two tabs at the bottom of the Workspace Window one for each pane view The Workspace Manager and Session Log can be hidden using the Auto hide function
123. values of Revenue per Unit depend on which state or alternative Product Strategy is in Arrowhead colors are covered in more detail in Sections 3 2 and 7 2 44 Syncopation Software Chapter 2 Building Decision Analysis Models Product Strategy In house E NPV License A NPV Figure 2 20 Model with Revenue per Unit Conditioned on Product Strategy DPL has also updated the Default Tree adding NPV as a get pay expression for the decision node Product Strategy A get pay expression tells DPL what value to get i e you receive or pay i e you pay out at each point in the tree Get pay expressions will be covered in more detail in Chapter 6 The get pay expression is displayed below the branch on which it occurs in the decision tree For now leave the get pay expression as is Value nodes are not displayed directly in the Decision Tree rather the names of value nodes appear in get pay expressions on branches in the decision tree Unlike a decision or chance node a value node is not an event which occurs in the sequence of the tree it is simply a calculated quantity 45 Chapter 2 Building Decision Analysis Models Syncopation Software You will now neaten up your model by bending the arc from Product Strategy to Revenue per Unit Select the influence arc with the left mouse button and without releasing the mouse button begin to drag the arc The mouse cursor will change to the arc bend cursor AA and a d
124. you added to the model with each of its branches and a blue triangle at the end of each branch As mentioned previously the blue triangle is DPL s representation for an endpoint As the model currently stands the Policy Tree has two endpoints one for the In house branch and one for the License branch The Policy Tree also displays the expected value or rolled back expected value at each point in the tree in square brackets The 105 1 next to Product Strategy is the expected value of the objective function for the complete model or tree Look further down the tree to the two rolled back expected values that appear next to each blue triangle and you see that the expected value at the endpoint for In house is 105 1 and the expected value for the endpoint at License is 93 7 As mentioned in Section 2 4 DPL maximizes the objective function of a model Therefore when confronted with a decision with two alternatives one of which has a rolled back expected value of 105 1 and the other a rolled back expected value of 93 7 DPL chooses the alternative with a rolled back expected value of 105 1 This can be seen in the Policy Tree not only by the 105 1 next to Product Strategy but also by the In house branch of Product Strategy being displayed with a thick branch DPL indicates the optimal alternative of a decision at each point in the Policy Tree by displaying the optimal alternative with a thick branch Recall that 105 1 is the same
125. you will need to add two additional nodes Create a new value node by dropping down the Model Node Add split button and selecting Value from the list Click near the top of the Influence Diagram to place the new node The Node Definition dialog appears Type License Fee for the node name Switch to the Data tab Type 0 75 for the node data Click OK to close the Node Definition dialog Click Model Influence Arc Add see Table 1 3 Draw an influence arc from License Fee to Revenue per Unit click on License Fee and then click on Revenue per Unit GUC 4 Note Another way to draw an influence arc is to hold down the Shift key click on the first node such as License Fee and then click on the second node such as Revenue per Unit You can also right click in the Influence diagram and select Add Arc from the context menu gt Repeat the process above adding a new value node called Price with data of 2 40 Syncopation Software Chapter 2 Building Decision Analysis Models Product Strategy In house m License lt Figure 2 16 Model with Price and License Fee Nodes You will now edit the definition of Revenue per Unit to incorporate the Product Strategy decision gt Double click the Revenue per Unit node The Node Definition dialog appears with the Data tab selected gt Click the Conditioning button The Conditioning dialog appears gt Click the checkbox next to Product Strategy
126. 0 30 Market Share Market Size No 174 0 Nominal 174 0 50 40 Market Size High 174 0 30 No 0 0 lt gt Figure 6 11 Policy Tree Expanded to 4 Levels 191 Chapter 6 Asymmetric Trees Syncopation Software When FDA Approval is No all three values on the branches of Market Share are the same 174 million This is the cost of developing the drug all the way to an FDA submission and being rejected If the drug is not approved Market Share Market Size and Pricing are all irrelevant The spreadsheet correctly handles this in calculating NPV in the disapproval scenarios As with the Develop decision this is another situation where asymmetry can make the model more efficient and intuitive Return to the Asymmetric model You can double click on its icon in the Workspace Manager window click on its tab in the document navigator or press Ctrl F12 Double click on FDA Approval in the decision tree Choose Asymmetric for the Outcome Grouping in the Tree Instance tab of the Node Definition dialog Click OK Drag the Market Share node and drop it on the Yes branch of FDA Approval Your model should look like Figure 6 12 Y U Nominal High Figure 6 12 Decision Tree with FDA Approval Asymmetric Unlike the decision to halt development an outcome of No for FDA Approval does not result in a value of zero The correct value is 174 million the cost of R amp D up to that point To correctly m
127. 00 Samples To continue the risk analysis you need to make the other key value drivers uncertain gt Switch back to the Risk Analysis model by pressing Ctrl F12 or double clicking its item in the Workspace Manager gt Change Trendline Revenue Growth Rate Share of at Risk Revenue Lost and RMC Lawsuit Settlement to continuous chance nodes You can also change node types via the ribbon by clicking the Model Node Change To split button DPL updates the default action of this button based on the last type of node you changed to 162 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models gt Enter the probability distributions and parameters as given in Table 5 2 Trendline Revenue Normal mean 0 08 standard deviation Growth Rate 0 01 Share of at Risk Revenue Triangular min 0 50 max 0 95 mode Lost 0 70 RMC Lawsuit Settlement Lognormal mu 4 0 sigma 1 0 Table 5 2 Probability Distribution Parameters for Continuous Chance Nodes Runa Monte Carlo Simulation with the same settings as Figure 5 19 Note DPL will issue a warning about deleting unsaved output results gt Click OK to the warning DPL brings up a new Risk Profile Histogram Finally you will examine the statistics available for the Risk Profile gt Click Data Distributions Statistics The Statistics dialog appears as shown in Figure 5 21 The Statistics dialog gives you a standard statistical description of
128. 1 Double click on the Policy Summary item in the Workspace Manager to activate the window The Policy Summary as shown in Figure 3 26 is displayed Product Strategy Price Change Sales 1 License does not occur 0 0 does not occur does not occur Cost per Unit Marketing Costs Price Nominal does not occur does not occur 0 0 0 does not occur Sales 2 does not occur Figure 3 26 Policy Summary 105 Chapter 3 Decision Analysis Results Syncopation Software The Policy Summary displays summary information about the optimal policy in the Endpoint Database For each node it displays the policy dependent probabilities for each state of the node at the end of the state branch A policy dependent probability is the probability weighted fraction of optimal scenarios in which the state either decision alternative or chance outcome occurs You will note that DPL also displays an additional state for each node called does not occur and displays the probability of this happening as well The does not occur state will have a non zero policy dependent probability if there are scenarios in which the node does not occur This could happen in an asymmetric tree a tree where not every node appears on every path Asymmetric trees are covered in Chapter 6 For initial decisions those that appear before any chance nodes in the decision tree the policy d
129. 15 DPL Programs and Code If you become comfortable with using DPL programs you may find them useful for quick what if analyses such as the change you just made to the discount rate You can always revert back to the original model by deleting the program from the Workspace Manager and re generating it from the original model The remaining sections of this chapter are intended for DPL users who need a more thorough understanding of DPL programs 15 4 DEFINING COMPONENTS IN THE DEFINITION SECTION The definition section is a list of definitions of events variables and constants This section can also include comments DPL is a very flexible language there are only a few rules for ordering punctuation and Spacing 15 4 1 Definition Order You may define the elements of a decision problem in any order you do not have to group all values together or have all decisions appear at the beginning of the definition section You can refer to previously defined events values or other elements in subsequent data definitions and in the sequence section The only limitation is that you cannot refer to a name before it is defined 15 4 2 Punctuation and Spacing DPL requires that statements that define elements of a decision problem end with a semicolon In contrast semicolons are not required in the second part of a DPL program where you define the sequence of events Definitions also have punctuation to separate the parts refer to
130. 2 growth Wages _phase_1 Wages phase Wages _salanes Location Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptione amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Initialization 097 E 7 Initialization C 8 E e Initialization 091 7 Initialization 091 0 Initialization 097 1 4 Assumptione amp Outpu Assumptions amp Outpu Assumptions amp Outpu CFIf0 14 0914 CFIS0 20 0 20 Assumptione amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu CFEC 0917 Type Value Value Value Value Value Value Value Value Value Value Value Value Value Value Value Formula Value Value Formula Formula Value Value Value Value Formula Count 4 st ost as l ss ss A a a a d ss i ss a a i i I M ss Rows L _ l l l l l st a l st l st a a u I et I I I st st DE Col Select All L Cancel size Prefie node ame as as a sa l l ss l St a OJ O i et it i et i tt i tt i ta st xX Sort by location Show ranges already linked Show ranges with formulas Create arrays bated on range name with sheet Figure A 3 Range Names Dialog for Creating Linked Values from Excel 440 Syncopat
131. 25 16 2 Spreadsheet Practices for Easier Conversion ccsccseeseceeeeeeeeeeeseeeeeeseeeaees 434 A Overview of Spreadsheet LINKING s csesesesesenenenensnsssseseeeeeeeeees 437 A 1 Types of Spreadsheet Links in a DPL Model cccecceceeeeseeeeeeeeeeeeeseeeeeees 437 Iv Table of Contents Syncopation Software AZ Calculation LINKS wascutuactndcaaaecadudusioducuartadenunasndusagutedsduauiecntadieweabediaaistatacaeess 438 AS TIGIAUZATION LINKS asienta iaeia Ea Ea 445 A 4 Managing Spreadsheet Links cccccceccseeeceeeeseesseessnecneseneseeeeceeesensaneseas 452 B System Requirements and Compatibility with Older Releases 455 C Keyboard ShortcutsS ssssssssssnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnan 457 D Glossary of DPL and Decision Analysis TermS sssesesesessesssee 459 INGEN arn a E O 479 Syncopation Software Chapter 1 Introduction 1 INTRODUCTION 1 1 WELCOME TO DPL S PROFESSIONAL Congratulations on your purchase of DPL 8 If you re a new user we re confident that DPL s power and flexibility will exceed your expectations If you ve used DPL in the past we think you ll find the new look and capabilities of DPL 8 significant improvements The DPL modeling environment was designed to give you the capabilities you need to focus on the problem at hand the task and not the tools This guide is an extension of that perspective teaching you essential DPL Skills in the context of simp
132. 253 9 1 Creating and Linking a One DimensionalValue NOCEe cccsceseeseeeeeeeeeeaees 254 9 2 Creating and Linking a Two Dimensional Value Node cccssceseeeseeeseeeeaes 271 10 Time Series Percentiles and Multiple Metrics ssssesesesesessssees 277 10 1 Time Series Percentiles ccccsccssccsecseeceeeseeceecaeeceesseeeeesseseeeseessesaeeeetsaeees 277 10 2 Risk Profiles with Multiple AttriDUteS ccccecesseseeeeeeeeeeeeseeeeeeseeeaeeseeeaees 295 10 3 Using Multidimensional Value Nodes in Get Pay Expressions cssscssceeees 300 11 Using the Endpoint Database ccccccssssessscsesenensnscseesseeseeeeeeeenes 305 11 1 Recording and Viewing Endpoints ccccccceeceeeeeeeeeeeeaeeeaeeeaeeeseeeseeeseenaees 306 TZ Playing ENDON seis inso oin oa erie moe EE PONE APAE 315 11 3 Reconnecting Endpoints to a Model s sssssssssssssssrsnrennrsnennrsnrnnrenrenrennnnnne 327 11 4 Exporting Endpoints isra minson an nA AAA Er Na 330 11 5 Importing ENA POMS a irnn ninani An A TATANA 334 12 Additional Decision Analysis Results cscscssssssssseeesesesssseeeeeeeees 341 12 1 Policy Tree Features sisi cataccadensdtecivasavadi cadicscadeiaviideatasisatedgaaaiaaaeeuaniae 341 12 2 Scenario Risk Profiles virsiin unninn annan nen een aa ara arai 348 12 3 Policy Summary FeatureS sessssssnsnssnnnsrrsnsnnrrrenrnrenrnnnnrnnnnrenennereneenennne 352 12 Option Value Charts caier iaria nA AENA EETA Eai 361 13
133. 5 4 Cost per Unit High 197 2 60 Marketing Cost alee High 169 9 40 Unit Sales License 93 7 Nominal 40 High 30 Price 56 6 Price 18 7 Price 0 3 Chapter 3 Decision Analysis Results Low 30 Nominal 40 High 30 Low 30 Nominal 40 High 30 Low 30 Nominal 40 High 30 44 4 44 4 52 0 52 0 74 8 74 8 6 5 65 14 1 14 1 36 9 36 9 12 4 12 4 4 8 4 8 17 9 17 9 Figure 3 4 Policy Tree Expanded to Endpoints Nodes in the Policy Tree are displayed using the same symbols as in the Decision Tree Decision nodes are yellow squares Discrete chance nodes are bright green circles Endpoints are blue triangles For a closer look at the data elements displayed in the Policy Tree and a description see Figure 3 5 Table 3 5 and Table 3 6 81 Chapter 3 Decision Analysis Results Syncopation Software Figure 3 5 Data Elements in Policy Tree Data element location Definition key Node Name 1 Rolled back expected The expected value of the objective values 2 function for the subtree headed by the node adjacent to the value For example in Figure 3 4 look at the top occurrence of Cost per Unit This is in the scenario where Product Strategy is In house and Unit Sales is low The expected value at this point in the tree is 18 3 displayed below the node name The probability of Cost per Unit being Nominal is 60 and being High
134. 6 230 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models 100 95 90 85 80 75 70 65 ray 60 S 55 a Y 50 None E E 40 O 35 30 25 20 15 10 5 ey ee SS a a ee a a SS 90 60 30 0 30 60 90 120 150 180 210 240 270 300 330 Figure 7 16 Risk Profiles for the Initial Decision Alternatives The Risk Profile shows how the seismic structure information has a risk mitigation effect The probability of an NPV outcome of 50 million or worse is 20 with the Core sample test and 50 without it The slight difference between the two alternatives at the higher end of the chart upper right simply reflects the cost of the test which reduces the value of the Core sample alternative slightly in the best case scenarios In the first implementation of the wildcat model above the Drill decision was being made without any information By adding the Test decision the Drill decision becomes a real option you have the ability but not the obligation to Drill after initially investing money in the testing decision The reason the expected value of the model increased after you added the Test decision is due to the learning in the imperfect information contained in the Core Sample test The Drill decision becomes an option to be exercised or not based on the results of the Core Sample outcome 231 Chapter 7 Conditioning and Learning in Decision Models Syncopation
135. 6 33 Mixed Outcome Group Definitions Click OK twice DPL has detached the Launch decision from the Key competitor outcome node Drag the Launch decision and its subtree to the blue endpoint for the Failure and Disappointment outcomes Release the left mouse button to attach the subtree to these two outcomes The decision model should now look like Figure 6 34 211 Chapter 6 Asymmetric Trees Syncopation Software Market Size Pricing Key competitor Launch outcome Low Low PV_Market Failure Yes Nominal Nominal Nominal Develop PV_Launch PV_Market lt q Disappointment No High High Yes PV_Market PV_Development Breakthrough No Figure 6 34 Decision Model with Mixed Outcome Grouping The tree structure now assumes that management will not launch the product if the key competitor has a breakthrough You will run the decision analysis to see how much this assumption reduces the value of the Develop decision gt Click Home Run Decision Analysis or press F10 gt Click OK for the warning 212 Syncopation Software Chapter 6 Asymmetric Trees The results should look like Figure 6 35 The value of the overall Develop decision has dropped to about 11 million a loss of 10 million Note that because this tree is small it should be easy to explain to management exactly why this is the case There is a 15 chance of ending up in the situation in which the competitor has a Breakthrough and if this h
136. 8 732 paths 138 Syncopation Software Chapter 4 Strategy Tables A Elast Elast Elast Base Var Base Var Base Var Festu U Strategy a Low Low Low Low Low Low Low Low Low Festure Led NPV Nomins Nominal Nominal Nominal Nominal Nominal Nominal Nominal Nominal Price Cuts NPV lt High High High High High High High High High Differentiate Product NPV w Figure 4 17 Revised Model with Strategy Node You can now run the model In this example you saw how to create a strategy table add a number of existing decisions to it create strategies and define the strategies in terms of the decision alternatives The Strategy Window interface also allows you to create decisions directly within the strategy table delete decisions remove decisions but maintain them in the Influence Diagram reorder the decisions in the strategy table by dragging and dropping the decisions in the decision columns and create and delete decision alternatives 139 Chapter 4 Strategy Tables Syncopation Software 140 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models 5 BUILDING AND ANALYZING MONTE CARLO SIMULATION MODELS This tutorial focuses on building an Excel linked DPL model for a financial risk analysis application It assumes familiarity with cash flow spreadsheets and with the analysis of uncertainty by Monte Carlo simulation This chapter dis
137. Appropriate to use for Event Tornado Event Tornado models with risk tolerance Can be difficult to explain Table 14 1 Summary of When to Use Each of DPL s Tornado Diagrams 389 Chapter 14 Sensitivity Analyses Syncopation Software Particularly good for Things to be aware of Rainbow Diagram Two way Rainbow Diagram Understanding the impact on the objective function of varying a single value across a number of settings determining if policy changes occur as a result of varying a single value across a number of settings Understanding how the interaction between two variable impacts the objective function determining if policy changes occur as a result of varying two values simultaneously Table 14 2 Sumary of When to Use Each of DPL s Rainbow Diagrams 390 Syncopation Software Chapter 15 DPL Programs and Code 15 DPL PROGRAMS AND CODE The DPL language DPL code is at the heart of DPL It provides a simple compact and precise way to define almost any decision analysis model using text rather than graphics You don t have to know anything about DPL code to use DPL If you are building large complex models or are converting spreadsheets however you may find it useful to look at DPL code and or work with it In a DPL Workspace DPL code is contained within DPL programs This chapter provides an overview of DPL programs and their components followed by a brief tutorial on converting D
138. Assumptions 0 32 Value ry rr Growth Assumptions E 18 Value AL Show ranges already linked Dhi St i et i st et l l _ l l L L Show ranges with formulas Dd cm Create arrays based on range size Prefix node name with sheet name Figure 9 16 Range Name Dialog for 2 Dimensional Array Double click the new node to edit it Switch the General tab of the Node Definition dialog name the node Yearly Growth Note that DPL has created it as a 2 dimensional array See Figure 9 17 2 3 Chapter 9 Multidimensional Value Node Syncopation Software Scalar value 1 dimensional array N values or series 2 dimensional array M x N values DPL variable name for expressions Yearly_Growth Figure 9 17 Node Definition Dialog for 2 Dimensional Array gt Switch to the Data tab of the Node Definition dialog Note DPL has created a 2 by 25 array and filled in the data from Excel See Figure 9 18 note en om O O O v Figure 9 18 Data Tab for 2 Dimensional Array with Values 274 Syncopation Software Chapter 9 Multidimensional Value Nodes gt Switch to the Links tab DPL has linked the node to the Yr_By_Yr_Growth range Click click OK Your model should now look like Figure 9 19 Yearly Growth is a value node with perpendicular vector arrows on it to indicate that it is a 2 dimensional node
139. Attributes Syncopation Software DPL will warn you that for fast sequence evaluation to work properly the conditions under which the halt function is executed must depend only on attributes or constant values gt Click Yes twice for the warnings The constraint has changed the optimal alternative If you were to conduct the Core sample test and then decide to drill the land constraint would be violated gt In the Policy Tree select the Seismic Structure node and drop down the Policy Expand split Select Expand Subtree from the list to expand that section of the tree gt Click the Zoom Full see Table 1 1 button or press Ctrl L See Figure 8 17 Drilling Costs Dri Yes 24 0 26 5 5 0 None 24 0 26 5 5 0 0 0 5 0 No 0 0 0 0 0 0 Drilling Costs Drill Yes 1000000000000 0 26 5 5 0 No 6 0 6 0 0 0 0 0 12 0 41 No 6 0 6 0 0 0 Test 24 0 26 5 5 0 i Drilling Costs F Seismic Structure Drill Yes 1000000000000 0 6 0 0 0 Core sample 6 0 6 0 0 0 Open 6 0 6 0 0 0 0 0 12 0 6 0 0 0 35 No 6 0 6 0 0 0 Drilling Costs Dri Yes 1000000000000 0 6 0 0 0 Closed 6 0 6 0 0 0 0 0 12 0 24 No 6 0 6 0 0 0 Figure 8 17 Policy Tree for Model with Constraint DPL cannot expand the subtree for Seismic Structure further beyond Drilling Costs because the halt function prevented DPL from continuing to evaluate that path If you double click on Drilling Costs the Policy Tree will not e
140. Chart Format Display gt Check EV CE Lines gt Check EV CE Values A line is added to each risk profile to show the expected value with its value to the left of it for each alternative See Figure 5 32 100 95 90 85 80 75 70 65 60 55 50 Yes 45 No 40 Cumulative Probability 35 30 25 20 j 15 10 5 0 200 0 200 400 600 1000 1200 1400 1600 1800 Figure 5 32 Risk Profile Chart Comparing Initial Decision Alternatives From this Risk Profile Chart you can see that the risk profile for the No alternative generally lies below and to the right of the risk profile for the Yes alternative This means that for almost any outcome value on the x axis there is a lower probability of the No alternative outcome being less than the x value than there is of the Yes alternative being less than the x value Also the No alternative has a higher expected value than the Yes alternative 175 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software For most decision makers the No alternative would be preferable to the Yes alternative It appears that the parent organization should not spin off Enlightify 5 9 PERFORMING A SENSITIVITY ANALYSIS In the previous section you concluded that it was better not to spin off Enlightify However you didn t test the assumption about the loss of synergy that would result from a spin off DPL
141. Conversion Options Display C Convert range names to lower case C Convert labels to DPL strings Generate display command procedure Code generation Wf Embed unnamed constants Embed unnamed formulas E Ignore unreferenced range names Suppress series generation Always prefix names with sheet name Unsupported spreadsheet functions 7 Suppress warnings 4 Replace with tna rather than 0 Defaults Figure 16 4 Conversion Options Dialog 428 Syncopation Software Chapter 16 Converting Spreadsheets to DPL Code You will rarely need to change any of the conversion options settings Table 16 1 through Table 16 3 describe what each option does Off Convert range Converts e g Growth_Rate to names to lower growth_rate Should not be used case with Convert to Calc Program from Model Links dialog Off Convert labels to Converts cells which contain DPL strings strings but are not used in calculations Generate display Generates statements which can command be used for advanced procedure debugging in the Command Window Table 16 1 Display Conversion Options definitions 429 Chapter 16 Converting Spreadsheets to DPL Code Embed unnamed constants Embed unnamed formulas Ignore unreferenced range names Suppress series generation Always prefix names with sheet name Cells which contain constants used in calculations will not normally be defined as DPL values DPL normally p
142. Creating Calculation Linked alues Mame Phase_1_ payment Phase_1_ success Phase costs Phase payment Phase success Phase_3_ costs Phase payment Phase 3 success prob fda approval Location Assumptions 0 51 Assumptions E DCFISE 8 PES Assumptions CSZ Assumptions l 22 DCFISE 5 V 9 Assumptions C53 Assumptions l 25 Probabilities 0912 Type Count Rows Col Value 1 Value 1 Formula 18 Value Value Formula Value Value Value l Chapter 6 Asymmetric Trees 1 SelectAll Cancel Sort by location Show ranges already linked Probabilities C 6 Value Probabilities 01 1 Value probs competitor outcome Probabilities C 14 Value Probabilities E 26 Value Probabilities C 22 9 Value Probabilities C 8 C Value Probabilities C 10 Value DCFIS0 11 DCFIfD 14 DCFI D 7 DCFI D 8 DCLFISD 9 DCFISD 32 DCFIfDF10 DCFIFE 10 7810 Formula Assumptions CS4 Value Assumptions 6 50 Value Assumptions C tS Value Assumptions CESS Value prob success phase prob success phases Show ranges with formulas probe_market_ share probe market size probs phases probs pricing PY Development PY Launch PY Phase_1 Py Phase Py Phase PY post_launch PY Registration Registration_costs Registration payne royalty pate Start_vear Create arrays bated on range Formula SIZ polite Prefix node name with sheet name Formula Formula
143. F Ey Smile h J Delete g Z l E E Hi Format 27 Fiter Select Calibri i Ul Alignment P Number Cells Editing E D H J This file is in the endpoint import format Comment lines may be included provided they begin with a double slash E Blank lines are ignored and may also be included Data needs to be comma separated The end of each line needs to be marked with a carriage return and line feed The first three non blank non comment lines must be the number of endpoints number of events and number of attributes in that order and in the format below The data for events must follow those three pieces of data with the name of the event first on the line and the states following Use the proper DPL identifiers for event names and states i e underscores in place of spaces or punctuation e g My Node becomes My_Node_ DPL identifiers are case sensitive The data for endpoints must follow the event data The beginning of the endpoint data must be marked by a line that starts with the phrase Endpoint Data The first non blank non comment line following the marker must contain the names of the events to indicate the order of the columns The data in the probability column may be left blank but the column must be included Number of endpoints 32112 Number of events 13 Number of attributes 1 Event Data Proceed Yes No Initial_Upstream_Inves Low High Pipeline Strategy Mkt_l
144. Figure 6 20 Market a Size Pricing FDA Approval Low Low P V_Market Yes Nominal Nominal Nominal Yes PV_Market High High PV_Market PV_Market Develop Figure 6 20 Get Pay Expression Revised by Drag Drop Method Now you need to place the other two import nodes in the appropriate decision tree paths as Get Pay expressions This time you will use the drag and drop method Development costs are only paid if the Develop decision is Yes launch costs are only paid if FDA Approval is Yes Select the PV Development value node in the Influence Diagram Drag it down to the Decision Tree pane The cursor changes to the add op get pay cursor get gt Drop it on the Yes branch of the Develop Node in the Decision Tree DPL adds PV Development to the Get Pay expression for this branch as shown in Figure 6 21 Pricing Low Low Develop PV_Market Nominal Nominal Nominal lt d Yes PV_Market PV_Development _ No a High High PV_Market PV_Market No q Figure 6 21 Get Pay Expression Revised by Drag Drop Method 200 Syncopation Software Chapter 6 Asymmetric Trees gt Repeat the drag drop procedure to place PV_Launch on the Yes branch of FDA Approval gt Select the No branch of FDA Approval gt Remove PV_Market from the Get Pay expression by deleting the expression in the Get Pay drop down and pressing enter Note If you drag and drop a value from the Influence Diagram pane to an existi
145. For a detailed description see On Line Help User Library A Dynamic Link Library DLL file which contains a set of functions or a program to be run during a DPL analysis and language that allows communication with DPL User Libraries are called from commands in the decision tree or program file User Library Function An external set of commands which perform a routine or calculation during a DPL analysis User Library Functions are Stored in a User Library which has a file extension of DLL Utiles The unit of measurement for the transformation performed by the Utility Function In a model incorporating risk tolerance DPL selects the policy which maximizes the expected utility i e greatest utile value The inverse of the utility function converts utiles to certain equivalents Utility Function A mathematical expression which captures the decision maker s Risk Attitude The Utility Function translates the outcomes of a value model into utiles In DPL the built in exponential utility function is defined as u x e x r where x is the value to be converted to utiles and r is the risk tolerance coefficient Alternatively you may specify a user defined utility function 4 6 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms V Value Correlations Chart A chart that displays the correlation coefficient between each value in the model including values associated with decision and chance events and
146. For s ana label ani4 po none Point hone Foint none Point none Point Run etting Evaluation method Full tree enumeration Number of intervals 500 Figure 10 10 Time Series Percentiles Dialog After Edits 288 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics The first time period in the Time Series Percentiles chart is labeled using the Initial time period Time periods are labeled sequentially after that DPL displays the current evaluation method and the number of intervals in the Run settings section In the Percentiles section you can specify the percentiles you would like included in the chart You may edit the existing three to change them to a different percentile type a number in a slot that says none to add one or select a number and press Clear to remove it If you d like two percentiles displayed as a range bar select a percentile and use the Type button to toggle which type it is You must specify both a range min and range max to display a range bar You may also set the type by right mouse clicking the Percentiles box and using the context menu Click the Change button to see the Run Options dialog See Figure 10 11 Evaluation method for Decision Analysis Fast sequence evaluation exact EV Full tree enumeration Discrete tree simulation approximate EV Samples for Monte Caro and discrete tree simulation 1000 Compile time optimizations Expres
147. HEN TO USE WHICH DPL SENSITIVITY OUTPUT Table 14 1 and Table 14 2 provide a summary of when it is most appropriate to use each of DPL s sensitivity outputs as well as when to be careful using a particular sensitivity output J88 Syncopation Software Chapter 14 Sensitivity Analyses Particularly good for Things to be aware of Value Tornado Determining which values in a deterministic model should be treated as uncertainties understanding the probabilistic impact of changes to values in probabilistic models Base Case Understanding the Can be difficult to interpret Tornado deterministic impact of for highly asymmetric chance events comparing models All conditioning the deterministic impact of and probabilities are changes to values to ignored Results are chance events sensitive to the definition of the Nominal state for each chance event Probabilistic Comparing to other results Takes longer to run Base Case since the base run i e Tornado vertical line will equal the expected value of the model as in a Risk Profile or a Policy Tree Ease of explanation Deterministic Understanding the Inappropriate to use for Event Tornado probabilistic contribution of models with risk tolerance each chance event to overall model uncertainty Understanding the impact of a chance event while preserving conditioning Useful for models with a lot of conditioning Probabilistic Same as Deterministic
148. High State ERE pial aa High Marketing_Costs Low Nominal High Cost_per_Unit Nominal Nominal High Unit_Sales Low Nominal High OK Cancel Figure 2 37 States for Base Case Tornado Diagram Dialog Note the default tornado type for the Home Sensitivity Tornado split button changes based upon the last tornado type you have run The default is now Base Case Tornado 67 Chapter 2 Building Decision Analysis Models Syncopation Software To run a Base Case Tornado Diagram you must tell DPL which state i e outcome is Low Nominal and High for each discrete chance node For three outcome discrete chance nodes this is usually straightforward If you have used the default discrete chance node settings provided by DPL the exercise is easy as is the case for Marketing Costs Price and Unit Sales Figure 2 37 indicates that the states specified to be Low Nominal and High for Marketing_Costs Price and Unit Sales are Low Nominal and High respectively Double click in the cell for the Low State of Marketing Costs You can see that there is a combo box that allows you to type or select the setting for the state of this chance node Use the arrow keys to move around the other cells for each state of each chance node leaving the settings as DPL has suggested You will also notice that since Cost per Unit is a two outcome discrete chance node one of its outcomes will have to be repeated for the Low Nominal or High state specific
149. However rather than continuing with the multi attribute version of that model you will use the original version in which the objective function NPV is the only attribute Therefore you do not need to complete the tutorials in Chapters 9 and 1 before completing this tutorial Select File Open Navigate to the Examples folder underneath where you installed DPL If you used the default location the path is C Program Files x86 Syncopation DPL8 Examples gt Select Energy Case da to open the Workspace as shown in Figure 11 1 306 Syncopation Software Chapter 11 Using the Endpoint Database da2 DPL Energy Case Energy Model sopi J FindRepisce J Objective Function Policy Summary J Endipowts mitai 1000 gt Conpiie Dension at T 7 T a a a 8 a N p n 22m nea E ppongsy ton AS Bru tot free rez Serer the ere mis quma taxa Rn f i were ne x P _ mm one O e i q j j ro va N a 2 SUU Deu sra Sone 4 Oo d 6 un O Tene 3 A i TA osc 5 j i4 ouni Sa m er i amp TREI Y eet VW q b at Energy Modei x Fos Heip press FI IE E Figure 11 1 Energy Case Model Workspace In order to record endpoints DPL uses the Full Tree Enumeration evaluation method If you click Endpoints in the Home Run group DPL will automatically set the evaluation method to Full Tree Enumeration Within the Home R
150. JPL C PROFESSIONAL USER MANUAL APATION DPL 8 Professional User Manual Syncopation Software Inc www syncopation com Copyright 2013 Syncopation Software Inc All rights reserved Printed in the United States of America Revised March 2013 Table of Contents Syncopation Software Table of Contents L UNTKOGUCUION sicisiiciciicecccienecscdes end cdeceewwansetenwasdensenaceceimedsceuswncastewnenedenss 1 1 1 Welcome to DPL 8 Professional ccccccseccseesseeceeeceesceeecseeseeesseessaesaneneeess 1 1 2 HOW TO Get Help with DPI 2 siccsicsteeavancvancisacvaaccsagnsiccseavantuasnaaabaasesontneiexaneaeds 4 Mic LEYS Cell AC OW sessd catccracee acct cetera Perce ead AEN 5 A JAS BIG TOUR OF DP asics crt tenets eee arat enews a eats ana aA AE aE 5 15 Whats Newin DPL G erranari ste wteawedawinawshavieawssavdeawssaxdsanctacvuntdinatenadiwauice 22 2 Building Decision Analysis Models ssssssssnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnn 23 2 1 Value Nodes and Spreadsheet LINKS cccccccsccseesseeeeeeseeeeeeeeeeeeeeseeeeeeeeeees 23 22 JAN S E acts tera E A ab tetera E art T A E T 32 2 3 Running a Modelinas a aa AA Tr REENE ER 35 24 Decision NOGOS miinan nnana AA AATE EE ranni 37 2 9 Value Tornado DlagralM cic cate cate cote cece cenecees casecous conecoua NANORANA ARARA 53 2 6 Discrete Chance Nodes sscsii sete cate cecc cede cesccesecavecess coveceea conecaeacautcautccesas teas tenet 59 2 7 Base Case Tornado DidGr
151. L8 Examples if the default location was used during installation DPL opens the Workspace as shown in Figure 13 1 365 Chapter 13 Risk Tolerance Syncopation Software eee ll rer 3 3 30 amp SS 8 d s OPL Rist Tolerance License vs In house prob ao Hons Mode View Oeste Help 5 ASSIOWS Delete O Select an J Ruk Profle A Poly Tree Lawei ai Monte Cario Samplet m H omedo Tene Sere Rename aa Copy 2 Find epilace Objective Function Poley Surrenary Endpoints Infial 1090 Compiie Decision Rainbow Option Value Orte amp Gor Je v Analysis gt DO init Dec ARs vorc Value Corretations Reitart 109 gt A tae Workspace Manager a RiakTolevarxceda gt License vs In house prob Session Log 9 OPL Decis Programmang Language Copyright 2003 2013 by Syrcopstan Sofware be a re n a Mcorperaing sofware ot icense to Syrcepsian Software hc Al nights reserved eww ANCON core Marketing suppon synccpation com Sales A US tobiree 1 O86 736 27S Product Per Unt Costs Outside the US 1 978 233 2509 Strategy Costs Professora Versen Low Low Perectuel In house Fi Normal Profit Release 2 00 00 r j f Sertai Nert er SYDVEOIIO1IG2001 T eq qo lt z License H High t 4 Profit i High Leeneed User dober Syncopeton Sofware he Hg High c Profit gt License wi In house prob x For Help press F Zoom 405 Full Prev ye
152. Low 1100 CapEx Nominal 2000 11 CapEx High Pipeline Capacity BCM 30 B50 E CapEx Low 1200 CapEx Nominal 22008 37 CapEx High excel Excel _ 1 Assumptions Actual Price chance Actual Price Low Nominal High 3 4 3 LOY Ti Actual Price Low 254 Actual Price Nominal oo Actual Price High excel Excel 1 Assumptions Actual Reserves chance Actual Reserves Low Nominal High 3 4 3 1000 Actual Reserves Low 3000 Actual Reserves Nominal 5000 Actual Reserves High chance Market 2 Political Stability Now Stable Unstable p chance Forecast Price Low Nominal High Actual Price 75 q 15 q Actual Price Low 125 q 75 q Actual Price Nominal leirta 15 q Actual _Price High Value Forecast Price Porecast Price 395 Chapter 15 DPL Programs and Code Syncopation Software Ly Forecast _Price Low 25 Forecast _Price Nominal Bo Forecast _Price High chance Market 2 Political Stability Future Stable Moderate Unstable Market 2 Political Stability Now is peg a Market 2 Political Stability Now Stable 05 1 85 Market 2 Political Stability Now Unstable excel Excel 1 Assumptions Market 2 Stability Future value Marker 2 Political Stabilvcy Future Marker 2 POLI treat stability Micke E Market 2 Political Stability Future Stable uy Market 2 Political Stability Future Moderate 205 Market 2 Political Stability Futu
153. Op Costs Research node has been added to the end of the longest subtree in the Decision Tree This causes the Get Pay expression to be moved from the branches of the Capex node the node previously at the end of the subtree to the branches of the newly added Ops Costs Research node You will reorder the tree so that the new node will be placed immediately before the Proceed decision Prior to reordering you will first move the Get Pay expression back to the branches of the Capex node using DPL s cut and paste feature Select the branches of the Ops Costs Research node in the Decision Tree and press Ctrl X to cut the Get Pay expression from the branches Select the branches of the Capex node and press Ctrl V to paste the Get Pay expression to the branches of the nodes Your Decision Tree should now look like Figure 11 12 320 Syncopation Software Chapter 11 Using the Endpoint Database Par STG Mas r Pat Sgeire Desenes SREE Herten Maiei 2 E E Low Low w iow Polish ume t 19 BOM Aree ae see aT 2 Tee Juya r f f Rs ev s a EA 2 i f Tne wr 2 P entan le renee ome t 20 80m _ bi uti f or sbi ves fj ms A ron ones PUES Em P MN pn i E N 0 os l Poise i ir Eer t i Ea j lil j tec 20 BoM Moderste l 36 At Jraleb Figure 11 12 Decision Tree with Cut and Pasted Get Pay Expression You are now ready to reorder the tree gt Click on
154. PL models to programs The last three sections of this chapter are intended mainly as a reference they provide more detailed documentation of the definition and sequence sections of programs as well as advanced techniques for using DPL programs 15 1 WHAT IS A DPL PROGRAM A DPL program is a text based description of the components of a decision analysis model A DPL program may contain a complete description of a decision analysis model with decision and chance events conditioning values and a sequence section or it only may contain a set of data definitions to be used in another program When you build an Influence Diagram and Decision Tree in the Model window and then run any form of analysis on it DPL converts the model to a DPL program written in DPL code It then compiles the program checking that the model is complete consistent and structurally sound Then it uses the analysis engine to evaluate the model and create the requested outputs 391 Chapter 15 DPL Programs and Code Syncopation Software You don t have to start with an Influence Diagram and Decision Tree If you prefer you can write a DPL program directly in a Program window Most people however prefer to use the graphics and data management tools in the Model window There are three situations in which you may wish to work with a DPL program directly e Converting spreadsheets DPL is able to convert Excel spreadsheets into DPL programs Models using con
155. Protect ariin Linge pape March J1 2013 Dress 2 09 00 Sete J Serial hanter Sr DPA M O20 8 Licenied Weer dukar iynbepeioe fonws Fi omieer of patter Ld Thee 2 tf cea x fone x forex foros fox sree J cra ecko x Cra x Sepai For Help press FI i Zoom 14 Ful Pey ot Figure 10 17 Workspace Manager with All Risk Profiles Double click on the Risk Profile dataset for CF_2014 to open it You can see that there is a substantial amount of uncertainty regarding the cash flow in 2014 due to the initial investments required the net cash flow will probably be negative See Figure 10 18 296 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics gt G Q 2 a v 2 amp 5 5 UO 39 50 45 40 35 30 a 20 15 10 2 0 CF_2014 Figure 10 18 Risk Profile for CF_2014 If you want to see detailed statistics on the probability distribution for cash flows in 2014 you can click Data Distribution Statistics For comparison purposes you might want to view multiple attributes on one Risk Profile chart With the CF_2014 chart still active click Chart Series Modify You can display up to 8 Risk Profile datasets on a single Risk Profile chart Leave the first series set to CF_2014 Use the drop down lists to fill in CF_2015 and CF_2016 for Series 2 and Series 3 in the Modify Data dialog See Figure 10 19 297 Chapter 10 Time Series Perce
156. RMS A Alternative A state of a decision node or decision event DPL chooses among alternatives during tree rollback to maximize or minimize the objective function The set of alternatives for a decision is the range of possible actions to take for that decision Always Gamble Tells DPL to always gamble on the outcome of a chance event overriding the Don t Gamble specification which is used during the creation of an Event Tornado Diagram Events with Always Gamble specified are never replaced by their expected values See Don t Gamble Arc See Influence Arc Array A one or two dimensional set of numbers Arrays can be used to store related numbers much as a look up table is used in a spreadsheet The syntax for a one dimensional array is arrayname column_subscript The syntax for a two dimensional array is arrayname row_subscript column_subscript Row and column subscripts start at zero i e the first element of a two dimensional array is arrayname 0 0 Array Formulas Operations in DPL ordinarily take scalars single values as operands and return scalars as results When an operand that is ordinarily required to be a scalar is replaced with an array the containing formula is said to be an array formula An array formula must be enclosed in the special symbols and the notation used by Microsoft Excel Array Subscripts Numbers or formulas used to reference the elements of an array For example arrayname 1
157. Risk Profile Dataset 100 Samples 164 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models DPL reuses the Expected Value dataset each time an analysis is run If you want to save a distribution you need to rename it You can see that rather than renaming Expected Value DPL made a copy of it For more information on saving Risk Profile Datasets see Section 3 6 The second item called Expected Value is the Risk Profile Chart Risk Profile Charts and Risk Profile Datasets are separate items in the Workspace Manager because one chart can display several datasets For more information on Risk Profile Dataset and Charts in the Workspace Manager see Section 3 2 Next you ll run simulations with larger numbers of samples Run a Monte Carlo Simulation with 1000 initial samples Rename the Expected Value Risk Profile Dataset Sf 1000 Samples gt Repeat the process specifying 10000 initial samples and renaming the Risk Profile Dataset 10000 samples The run with 10000 samples may take a minute or two You can take the opportunity to switch over to Excel and watch the numbers change Excel needs to recalculate Total Value 10000 times Now you have three Risk Profile Datasets which you can display in a single Risk Profile Chart gt In the Workspace Manager rename the Expected Value Risk Profile Chart to Sample Size Comparison gt Double click the Sample Size Co
158. Screen The Node Definition dialog returns to normal dialog mode gt Click OK Your model should now look something like Figure 3 21 99 Chapter 3 Decision Analysis Results Syncopation Software Sales Marketing Price Sales Product Price 1 Cost Costs 2 Strategy Change per Urut _Low Low ies Low in house Raise Nominal f Sales 2 Nominal Nominal Aaa Nominal Lal License C Keep Same eme High CEOP a gh Nomina High High High High NPV Sales _2 Figure 3 21 Model After Editing Sales 2 DPL has indicated the changes that you have just made with colored arcs A green arc indicates that the influencing node i e Sales 1 called the predecessor conditions the probabilities of the influenced node i e Sales 2 called the successor As you have seen previously a blue arc indicates that the influencing node i e Price Change conditions the values of the influenced node i e Sales 2 If the predecessor conditions both the probabilities and the values of the successor DPL indicates this with a light orange arc Arc colors are covered in more detail in Section 7 2 Note that DPL continues to update the Default Tree as you make changes in the Influence Diagram The Default Tree is not yet correct but ignore it for now There are several more changes to make to the Influence Diagram 100 Syncopation Software Chapter 3 Decision Analysis Results We need to create a value node to connect t
159. Series2 from 1 to 10 Seriesl S 2 With these definitions Series1 2 equals 3 Series2 2 equals 6 Series1 10 equals 1023 and Series2 10 equals 2046 The first definition in the example indicates that the variable Series2 is a series rather than an array You can refer to series in three ways Assume the series Cash_Flows is defined as below series Cash Flow from 2014 10000 from 2010 tO 20238 Cash Elow lesb bedt You may refer to Cash_Flow as follows e With a subscript referring to a single element e g value CF2014 Cash_Flow 2014 CF2014 is 10000 e Without a subscript referring to the entire series e g value CFTotal sum Cash_Flow 407 Chapter 15 DPL Programs and Code Syncopation Software CFTotal is 159374 25 e With a subrange specification referring to a selected subset of the elements in the series e g value CFSubTotal sum Cash_Flow 2017 2020 CFSubtotal is 61771 71 For more information on using series refer to the online Help 15 4 10 Arrays You can define one or two dimensional arrays of numbers in DPL The following statement defines a one dimensional array in DPL array A 1 2 3 The first element of this array is referred to as A 0 and it equals 1 The second element is A 1 and it equals 2 and so on Commas separate elements in a one dimensional array The following statement defines a two dimensional array in DPL array B ly 2 3 A S G
160. Software Conditioning Land Disturbed The values for this node depend on Amount of Oil J Drill Drilling Costs E Oil Price Seismic Structure Number of value expressions required 4 Cancel Figure 8 3 Conditioning for Land Disturbed gt Click OK Enter data for Land Disturbed as indicated in Table 8 1 Your value input tree should look like Figure 8 4 when you are done Land Disturbed Value wee Table 8 1 Data for Land Disturbed Value Node Value 238 Syncopation Software Chapter 8 Objective Function and Multiple Attributes Node Definition Land Disturbed oes ae General Conditioning Print Page Setup Copy FullScreen Wheel Reorder QQ CE vii Links Land Disturbed Yes Test Core sample Drill Core sample Figure 8 4 Node Definition Dialog for Land Disturbed gt Click OK Note don t be concerned that Drill comes before Test in the data input tree The actual order of the nodes is dictated by the Decision Tree DPL creates new influence arcs from Drill and Test to Land Disturbed You may want to rearrange the nodes in your Influence Diagram for better readability Next you ll create a new value node for the weighting of disturbing land gt Click Model Node Add to add another value node to the model Place the node to the right of Land Disturbed gt Name it Land Weight gt Click the Data
161. Switch to the General tab of the Node Definition dialog See Figure 2 6 28 Syncopation Software Chapter 2 Building Decision Analysis Models Scalar value 1 dimensional array N values or series 2 dimensional array M x N values DPL variable name for expressions Cost_per_Unit Figure 2 6 Node Definition General Tab gt Switch to the Links tab of the Node Definition dialog See Figure 2 7 General Calculation links These links are recalculated throughout the run They may be either inputs used in the linked spreadsheets calculations g unit sales or calculated results from Data the spreadsheet e 9 NPV None local DPL Program Microsoft Excel Workbook C Program Files 6 lt 86 Syncopation DPL8 Examples New Product Strategy ds Browse Sheet Cell Drivers Cost_per_Unit GotoCell Cell Names y This is an export node Its value will be sent to Excel each time it changes during the run Initialization links These links are for connecting to a spreadsheet or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at the start of each run None local or DPL program Microsoft Excel Figure 2 7 Node Definition Links Tab 29 Chapter 2 Building Decision Analysis Models Syncopation Software The Links tab contains information about the links b
162. Syncopation Software Chapter 16 Converting Spreadsheets to DPL Code If you forget to save a version of the linked model and you follow the steps above to convert to a DPL program there is an easy way to undo this You will undo the conversion now and re link your DPL model to your spreadsheet Switch back to the cash flow decision model Select Model Links Options gt In the DPL Calculation Program section click the Revert to Linked button You will need to indicate the location and name of the linked spreadsheet DPL will re link your model as it was originally gt Click OK for the warning gt Locate the Cash Flow1 xls file when prompted by DPL After you revert to the linked model you will see in the Workspace Manager that the converted spreadsheet program has been moved to the Unreferenced programs section because it is no longer used by any of the models in the Workspace Manager See Figure 16 6 S Unreferenced programs Figure 16 6 Workspace Manager with Unreferenced Programs If you prefer you can delete an unreferenced program from the Workspace by clicking on it and deleting it DPL will ask you to confirm that you wish to delete the item 433 Chapter 16 Converting Spreadsheets to DPL Code Syncopation Software 16 2 SPREADSHEET PRACTICES FOR EASIER CONVERSION While DPL supports the functions most commonly used in financial spreadsheets Excel s functionality is vast so it i
163. The Node Definition dialog appears with the General tab selected You will now remove one of the chance outcomes that DPL provides by default On the General tab of the Node Definition dialog select the Low outcome in the Outcomes list box and click the Delete button Click OK The Node Definition dialog should now look like Figure 2 34 Name and settings Data Cost per Unit Links DPL variable name for expressions Cost_per_Unit Outcomes Nominal Defaut High Node Definition Cost per Unit Hide _ Separate probability and value data input trees Comments Add Insert Rename F2 Default Print All Comments Copy to Clipboard OK Cancel Help Figure 2 34 Node Definition Dialog General Tab for Cost per Unit Select the Data tab of the Node Definition dialog Because Cost per Unit is conditioned by the Product Strategy decision there are two sets of probabilities and values one for In house and one for License See Figure 2 35 No probabilities are given for the Nominal and High outcomes The default probabilities are only given for chance nodes with the same number of outcomes as specified by the defaults in File Options General However the Nominal and High value data that you specified for the In house alternative when you ran the Value Tornado are filled in 63 Chapter 2 Building Decision Analysis Models Syncopation Software Node Definitio
164. The probability weighted average of all possible outcomes Exponential Distribution A named distribution supported by DPL For a detailed description see On Line Help Export Node Export Variable A node linked to a specific cell in a spreadsheet to provide input for the spreadsheet model DPL exports data to the cell Data exported by DPL will overwrite any information contained in the cell Fast Sequence Evaluation The default method of calculation DPL uses for an analysis This is the fastest evaluation method that calculates the exact expected value Certain model structures are not suitable for Fast Sequence Evaluation and should instead be analyzed using Enumerate Full Tree Fat Policy The capability of DPL to display in the Policy Tree the expected values of all attributes in addition to the objective function at any point in the tree Frequency Histogram A view option in the Risk Profile Chart It plots the probability of occurrence for intervals of outcome values as a vertical bar Full Tree Enumeration An evaluation method that evaluates each path of the tree completely This is in contrast to Fast Sequence Evaluation 466 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms G Gamble An evaluation action taken by DPL for a chance event when evaluating a decision tree When DPL gambles on a chance event it creates the alternative outcomes of the chance event assigns them a probabil
165. Unit Sales Revenue per Unit Marketing Costs and NPV as indicated in Figure 3 11 One way to quickly re arrange the Influence Diagram is to select multiple nodes at once While holding down the Ctrl key click on the Unit Sales Marketing Costs Cost per Unit and Product Strategy nodes All these nodes turn magenta because they are selected as a group Use the mouse to drag this group of nodes to the left freeing up space in the center of the diagram You may have to adapt these steps slightly depending on how your diagram is drawn Figure 3 11 Re arranged Influence Diagram 91 Chapter 3 Decision Analysis Results Syncopation Software gt Select Model Node Add Decision from the ribbon to add the new decision node Click somewhere in the space you created The Node Definition dialog appears with the General tab selected cy gt Name the node Price Change cy Rename the alternatives to Raise and Keep same The General tab should look like Figure 3 12 Node Definition Price Change Generat Name and settings Price Hide _ Change Links DPL variable name for expressions Price_Change Altematives Raise Default Add Insert Delete Rename F2 Default Comments Print All Comments Copy to Clipboard ox Ca Ha Figure 3 12 Node Definition Dialog General Tab for Price Change Decision 92 Syncopation Software Chapter 3 Decision Analysis Result
166. You will zoom in on a portion of the Policy Tree yyy yyy cas fas 5 D g me O me en 6 7 Oo 0 4 ig m fs and OR mad rar mt q Figure 12 5 Filtered Policy Tree with Expanded Subtree 345 Chapter 12 Advanced Decision Analysis Results Syncopation Software gt While holding Ctrl Shift left mouse click above the Market Size node at the end of the Nominal branch of Market Share and without letting go drag down and to the right to create a rectangle around the endpoints of the expand subtree as shown in Figure 12 6 o vna Sak lt A 0 lt lt q ee a g 7e rs i ar 4 m lt q iji n lt wate E lt lt 2 gt ma aoe lt o lt lt j Figure 12 6 Using the Mouse to Zoom a Portion of the Policy Tree Release the mouse button DPL zooms to the portion of the Policy Tree you selected as shown in Figure 12 7 346 Syncopation Software Chapter 12 Advanced Decision Analysis Results Yes 400 4 2 122 9 Pricing Nominal 172 7 Nominal 162 5 40 469 7 40 High 319 7 Market Size 30 616 7 Nominal 126 9 Low 99 0 40 Pricing 30 395 9 High 329 5 Nominal 343 4 30 40 640 3 High 541 6 30 838 5 Low 12 3 Pricing 30 Market Share Low 194 8 Nominal 205 8 lt q Yes 221 0 30 40 502 7 122 9 High 30 659 6 Low 411 1 Market Size Pricing 30 708 1 High 741 9
167. able DPL has included the decisions you selected previously If you had not selected decisions DPL would have created two default decisions called Decisioni and Decision2 which would be displayed in the columns to the right of the strategy column In each decision column DPL displays the decision alternatives for the decision that heads the column Under each decision alternative in each decision column DPL displays the strategy icon that applies to that alternative As indicated by Figure 4 6 both Strategy 1 and Strategy2 are defined by the first decision alternative of all the decisions in the strategy table Therefore Strategy1 and Strategy2 are initially identical The Strategy Window has a number of buttons across the top Most of the commands that you will use regularly can be found by clicking the Commands button 129 Chapter 4 Strategy Tables Syncopation Software or by right mouse clicking in the Strategy Window to bring up the Strategy Window context menu as show in Figure 4 7 Depending on what you have selected in the Strategy Window different menu items will be enabled in the context menu Full Screen Add Strategy Add New Decision Add Alternative Add Existing Decision Break Out Decision Rename Delete Figure 4 7 Context Menu for Strategy Window You will now define some distinct strategies To start you will rename the existing strategies Select Strategy1 It will be displayed in magenta when i
168. accomplished by Shift double clicking the node Expand to Level Brings up the Expand Policy Tree dialog in which you can specify the level to which you i want the subtree expanded The Policy Tree is expanded from the node selected Expand Subtree Expands the entire subtree in the Policy et Tree from the selected node os P Expand All Expands the entire Policy Tree so that all nodes are expanded I gt Table 3 4 Policy Tree Expand Commands gt With the node still selected drop down the Expand split button and select Expand to Level The Expand Policy Tree dialog appears 79 Chapter 3 Decision Analysis Results Syncopation Software gt Enter 4 in the Expand to Level edit box in the dialog Leave Zoom full after expanding checked gt Click OK The Policy Tree should now look similar to Figure 3 3 Marketing Costs f Nominal 24 4 Cost per Unit 60 Low 18 3 Marketing Costs 30 f 40 Marketing Costs i Nominal 119 9 Unit Sales Cost per Unit 60 In house 107 7 Nominal 107 7 Product Strategy ppe Marketing Costs 107 7 iiis High 89 6 40 Marketing Costs R Nominal 215 4 Cost per Unit 60 i 7 2 nigh Meats Marketing Costs 30 f High 169 9 O 40 Unit Sales License 93 7 Figure 3 3 Policy Tree with In house Subtree Expanded 3 Levels Note that once the subtree has been expanded and the discrete chance node it exp
169. ace in the decision tree Allows the subtree to be drawn once and performed several times Poisson Distribution A named distribution supported by DPL For a detailed description see On Line Help Policy Summary An output generated by DPL which displays each event and its associated distribution under the optimal policy resulting from the analysis You can also use the Policy Summary to compare the optimal policy distributions to distributions associated with a range of the Risk Profile or to when a decision or chance is in a particular state Policy Tree A graphical representation of the optimal policy for every decision in a model given the outcomes of all the uncertainties in the model Shows all possible paths through a decision tree and indicates the value of all expressions in the model the probabilities associated with each chance event outcome the rollback values EV or CE for each node and the optimal policy choices for each decision Prob Function A function that returns the probability of an event at any point in an analysis at which the state of the event is unknown This function is useful in decision problems in which the probability of an event is itself a major consideration in making decisions Probabilistic Base Case Tornado Diagram Used with probabilistic models a diagram generated by evaluating the outcome of a model when each chance event is individually set to a high and a low state while all other chance
170. ach variable and the resulting objective function i e NPV values gt Check the EV CE Values checkbox in Chart Format Display The tornado is updated as shown in Figure 2 30 Value Tornado Cost per Uret Product Strategy In Pouse 1 937 pai 0 0 105 1 Expected Value E Low Policy Change Ei Policy same as Nominal I High Policy Change Figure 2 30 Value Tornado with Values Displayed DPL displays the low and high settings for each variable as well as the expected value of the objective function when the variable is at its low and high setting The variables low or high settings appear before the character and the expected values appear after In addition DPL displays the expected value for the Base Result at the bottom of the vertical line in the Value Tornado If a variable has no impact on the objective function as it is changed from its low to high setting DPL displays a zero width bar and no labels for that variable as is the case for License Fee Yd Chapter 2 Building Decision Analysis Models Syncopation Software The Value Tornado indicates that four variables result in a change in policy if they are at either their low or high setting These variables are called decision sensitive In this sample model there is only one decision with two alternatives so we know that a policy change implies that rather than producing the product In house the optimal alternative is to License it For more informa
171. al support for the most up to date information on spreadsheet conversion and Supported functions 455 Chapter 16 Converting Spreadsheets to DPL Code Syncopation Software 436 Syncopation Software Appendix A Overview of Spreadsheet Linking A OVERVIEW OF SPREADSHEET LINKING DPL provides a complete set of features for linking to Excel spreadsheets Most DPL analyses involve spreadsheet linking at some point This appendix summarizes DPL s main spreadsheet linking features and gives suggestions on when and why to use each feature A 1 TYPES OF SPREADSHEET LINKS IN A DPL MODEL Calculation Links SH are appropriate when the value measures being used for the analysis are calculated in a spreadsheet model usually a financial model producing NPV IRR etc Most DPL analyses employ calculation links and there are several ways to set them up Initialization Links es w i are used when it is more convenient to keep the numbers initializing DPL nodes probabilities and values in a spreadsheet rather than within DPL node data With initialization links you can design your DPL model as a template to be used with several input data spreadsheets A DPL model can have either or both types of links DPL s linking features are intended to work with input and output cells that are named ranges in Excel In general when developing an Excel spreadsheet that you intend to link to DPL you should make a habit of creating range names
172. al Option See Downstream Decision Reduction A DPL tool used to reduce an output probability distribution Risk Profile to a single chance event The reduced event will be described by a discrete probability distribution with a user specified number of states The output in the form of DPL code is written to the Session Log or pasted as a discrete chance node to a specified model Risk Attitude Decision maker s willingness to gamble relative to the expected value One who is not willing to gamble in the face of a positive expected value is risk averse while one who is willing to gamble in the face of a negative expected value is risk seeking One who makes decisions based on expected value is risk neutral Risk Profile The probability distribution for a specified policy comprised of the entire range of outcomes that are possible given the policy Risk Profile Chart A window displaying a probability distribution for a specified policy A Risk Profile Chart displaying a cumulative probability distribution of the optimal policy is a typical output of a DPL decision analysis Risk Profile Dataset The underlying probability distribution data for a specified policy which is typically displayed in a Risk Profile Chart Risk Profile Datasets may be saved in a Workspace without being displayed in a Risk Profile Chart 4 2 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms Risk Tolerance Coefficient Measur
173. al scenarios These event states are positively correlated with the decision alternative used for the filter For example in Figure 12 14 you can see that Seismic Structure Open and Seismic Structure Closed both have red bars that are longer than the blue bars This means that both these two outcomes are more likely given that the 356 Syncopation Software Chapter 12 Advanced Decision Analysis Results Yes alternative to Drill is optimal This tells you that these outcomes are positively correlated with the decision to drill You can also see that oil price is more likely to be nominal or high in scenarios where drilling is the optimal decision alternative Conversely chance outcomes or decision alternatives that have a red bar that is shorter than their blue bar are less likely in the filtered scenarios than they are in the full set of optimal scenarios These event states are negatively correlated with the decision alternative used for the filter For example in Figure 12 14 you can see that Seismic Structure No has no red bar at all This means that this outcome never occurs given that the Yes alternative to Drill is optimal This information can be used to help develop a decision rule i e if Seismic Structure is No then do not drill You can also see that oil price is less likely to be low in scenarios where drilling is the optimal decision alternative in fact it is 50 less likely 15 vs 30 Double click the Risk Profile Ch
174. alculated results from the Data spreadsheet e 9 NPV None local DPL Program Microsoft Excel Workbook Cash How with Decision xs Browse Sheet Cell Assumptions amp Output Marketing_phase_2 Go to Cell Cell Names y This is an export node Its value will be sent to Excel each time it changes during the run Paste Link Initialization links These links are for connecting to a spreadsheet database or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at the start of each run C None local or DPL program Microsoft Excel _ Database Workbook Browse Same as calc links workbook Go to Model Links ox Cones tb Figure A 7 Node Definition Initialization Links In the Excel spreadsheet make sure that you have named a range containing the data you wish to initialize the node In Figure A 8 the range Mkt_ph2_probs has been defined and it contains the probabilities 0 3 0 4 0 3 For unconditioned nodes the range can be a row or a column 447 Appendix A Overview of Spreadsheet Linking Syncopation Software Xl Wd I Cash Flow with Decision xis Compatibility Mode Microsoft Excel File Home Insert Page Layout Formulas Data Review View Developer DPL a o we 8 B amp PET General v Eg Conditional Formatting 3 Insert E AF Ba H7 7 gt Formatas Table g9 Delet
175. alog appears similar to Figure 3 32 but blank gt Press the Select button to select a variable for the dialog Select License_Fee in the Value for sensitivity drop down list as shown in Figure 3 31 Select Value for Rainbow Diagram A Value for sensitivity License_Fee kd Current yalue 0 75 Cancel Figure 3 31 Select Value for Rainbow Diagram dialog Click OK In the Run Rainbow Diagram dialog specify the From and To values to be 0 7 and 0 9 respectively 4 gt Set the Step size to be 0 025 The Run Rainbow Diagram dialog should now look like Figure 3 32 113 Chapter 3 Decision Analysis Results Syncopation Software Rainbow Diagram License vs In house downstream Vanable Variable name License Fee Select Current value 0 75 From a To a O Number of steps 9 Step size Evaluation method Full tree enumeration Change Cancel Help Figure 3 32 Run Rainbow Diagram for License Fee gt Click OK to run the Rainbow Diagram A Rainbow Diagram displays the value of the model s objective function on the y axis and the values of the selected variable on the x axis The diagram shows how the objective function value changes as the variable setting changes across the specified range The Rainbow Diagram for License Fee should look like Figure 3 33 The diagram indicates that as License Fee increases from 0 7 to 0 85 there is no change in the
176. alog box launcher The various Model options are discussed throughout the manual Select File Options to access the DPL and Workspace level options The Options dialog appears with the General tab active as shown in Figure 1 5 General Display F Show Tips V Restore window position on startup Workspace V Colored amowheads arcs Lines to clear when max exceeded 100 Outputs Max number of lines to display in log 1000 Undo levels Graphic models 30 100 max ODPL code 100 1000 max Default state names Outcome names probs for new chance nodes Alttemative names for new decision nodes Name Probability Name Low outcome Low 3 First altemative Yes Middle outcome Nominal t Second altemative No High outcome High After node name change Replace old with new in node data and get pays Dont replace Figure 1 5 Options Dialog with General Tab Selected 20 Syncopation Software Chapter 1 Introduction Make sure that Show Tips is checked Select the Outputs tab Make sure that Decimal places is set to 1 Make sure that Display probabilities as percentages is checked LUULU Make sure that Decimal places for probabilities is set to 0 The Outputs tab should look like Figure 1 6 Options General General number formatting Decimal places eg 123 5 Wworbaiare Output scaling 0 display 1 0 as 1 0 Zero equivalent 1 display 9 Se 013 as 0 Probabilities Display probabilities as percentages Dec
177. aluating models based on Discrete Tree Simulation In Modified Monte Carlo samples are first allocated to branches of chance events using the branch probabilities rather than completely random sampling Monte Carlo Simulation Method of evaluating models utilizing random sampling of continuous probability distributions chance nodes with a set sample size See also Discrete Tree Simulation Multi attribute Utility Analysis abbreviated MUA or MAU A method for making decisions where the value function depends on more than one factor such as cost and schedule DPL allows you to track multiple attributes using independent value expressions and combine them in a single utility function Multidimensional Value Nodes Value nodes that refer to one or two dimensional arrays or a series in a DPL model 469 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software N Name A string of characters used to identify a node string or constant Nodes will be referenced by their names when used as variables Named Distributions Pre defined probability distributions which can be assigned to chance events in DPL based on input for one or more parameters DPL supports 21 different Named Distributions such as the Normal Distribution and the Beta Distribution Negative Binomial Distribution A named distribution supported by DPL For a detailed description see On Line Help Node A graphical representation of a decision val
178. alue of 0 The correct definition is value Profit Revenues Costs Default Case Notation DPL provides another way to specify the probabilities for a chance event that depends on other events This approach can simplify the task of specifying the probabilities when the conditioning events have many combinations of states For example rather than specifying each probability distribution individually chance Atolls S52 S3 S4 1 2 3 chance B Yes No A 1 1 1 5 you can define a default distribution and then define only the exceptions chance Ay oly S27 S3 S4p fulg 2y ahs chance B Yes No A default 1 case A S4 5 If there are multiple conditioning events you can use logical OR and logical AND amp amp to specify combinations of events for the cases As with the probability distributions in chance event definitions you can save time and effort in a value definition by defining one value as the default for the value and identifying particular cases for other definitions 405 Chapter 15 DPL Programs and Code Syncopation Software 15 4 9 Series A series is a one dimensional vector of numbers whose elements are initialized in groups or intervals series Sales from 2014 0 from 2015 100000 from 2017 to 2020 150000 In this example the series is called Sales Sales contains three intervals one that has one element corresponding to the year 2014 one
179. alue xls Creating Calculation Linked alues Name Location Type Count Rows Col aa Include Market 2 No Assumptions 0469 Formula Select All Include Market Yes Assumptionsl C 70 Formula Initial Marketing Spend Assumptions C 33 Value Mkt 1 All Years Assumptions E 15 Formula Mkt 1_ Growth Assumptions C 9 Value Mkt Growth Assumptions 0910 Value OpCosts_Growth_1 Assumptions C 26 Value OpCosts_ Growth Assumptions C 2 7 Value Price_Growth_1 Assumptions l 24 Value Price_Growth_2 Assumptions C 25 Value Start year Assumptions C 29 Value Ten year CF Assumptions E 42 Formula Uptront_Cost_Escalation Assumptione l s2 Value Yr bp r Growth Assumptione tE 18 Value Cancel Sort by location Show ranges already linked Show ranges with formulas ae l l l l l l l l 1 2 Ph cm Create arrays based on range ize Prefix node name with sheet name Figure 9 4 Range Name Dialog for Ten Year CF for 1 Dimensional Array Click OK DPL places the node off to right You may move it to near the Market Size 1 node if you d like Double click the new node to examine its definition gt Switch to the General tab Note that DPL has created a one dimensional array or series Change the name to Ten Year Cash Flow 2359 Chapter 9 Multidimensional Value Node Syncopation Software Node Definition Ten Year Cash Flow Name and settings
180. alues and probabilities associated with the model and then aggregates the endpoints into a user determined number of intervals for display The number of intervals used can significantly affect the appearance of the Risk Profile chart The number of intervals has no impact on the expected value of the Risk Profile 86 Syncopation Software Chapter 3 Decision Analysis Results The precision of the percentiles from a Risk Profile chart is affected by the number of intervals In the example in the previous paragraph 40 is the 20 percentile The 10 percentile is the x axis value found by drawing a horizontal line from 10 on the y axis to where it intersects the curve and then drawing a vertical line drawn from the intersection to x axis If you are interested in percentiles and require more precision you should set the number of intervals to be larger A larger number of intervals requires more memory and has a slight performance impact The number of intervals can be changed by selecting Home Run Options Double click on the active model or press Ctrl F12 Click Home Run Options to launch the Run Options dialog Under the Risk Profile section you should see that the Number of intervals is 500 Click Cancel In general 250 to 500 is usually a reasonable setting for most models Note in the model as it currently stands there are only 108 paths in the tree half of which are in the optimal policy and d
181. ams ccceccsscesseeeeeeseeeseeesaeeeaeeeseeeaeeeaeeeseeeseesansaes 67 2 8 Initial Decision Alternatives Tornado DiaQram ccccccsseesseeseeesseeeeeeaeeeaeeees 70 3 Analyzing Decision Analysis Models cscscscscsssseseeeecesesseeeeeeeeeees 73 Sil PONEV TEE T naaa anaia aaa aa a a 78 322 RISK PO ile read e i ons gate vons i E A AEAEE aaa 85 3 3 Poley SUMMA M earair EAE EA 91 3 4 Expected Value of Perfect Information Control sesessssssenensrnnnenensrrnrenenne 107 2I NaS COrreladONS auan a E E EEE EEEE EEA 110 36 SAVING OPUS arara iae EEAS EENE EEEN ENE EAE EEEE 112 3 7 Rainbow Diagram on a Value uu eecceeeeeeeee eee eeeeeeeeeaeeeeeeeae essa eesaeesaeesaeeneeea 113 3 8 Rainbow Diagram on a Probability s sssssesssssenssnnnnnsnnnrnnrennnnnnnnnnnnennenrnnnns 117 3 9 Probabilistic Base Case Tornado Diagram ssssssssssssrssrrsresrrnnnnnrnrrnnrennennns 122 A Strategy Tables cscri oinnia anana aaaea p ANa aana NaRa 125 5 Building and Analyzing Monte Carlo Simulation Models 141 SL The Excel Cash Fow Model iscasena E 142 5 2 Creating ENKEA Vale Seana aaa aa a ra E Eaa 145 5 3 Arranging the Influence Diagram sssssssssssssrsrrsrrnrrsnrrnrrnrrnrrenreneenrnnnnnnnnns 150 5 4 Running a Value Tornado Diagram sssssssssssssssssnssnnsnnnrrnnrenrsnnnnnnnnennennnnne 152 5 5 Using the Workspace Manager s sssssssssssssrssrsrrsnnnnnnnnrnnrnnrnnrenrenennrnnenrennns 157 5 6 Running a Monte Carlo Simula
182. an Endpoint Database Playing endpoints works whenever DPL can re use the Endpoint Database to get the results you need The Endpoint Database stores all the value model information that is needed to roll back the Decision Tree Typically calculating this value model information particularly if this information comes from an Excel spreadsheet that needs to be re calculated is the most time consuming part of a DPL run When you play endpoints DPL only needs to look up values rather than recalculate them which results in much quicker runtime If you change your value model i e add a value node which contributes to an attribute or the objective function after you record endpoints you will not be able to play endpoints If you have not changed your value model playing endpoints is usually possible The exception is when you have changed the structure of the model such that there are new endpoints This can occur if you record endpoints for an asymmetric tree and then reorder nodes Reordering nodes in a symmetric tree changing probabilities adding conditioning events branch controlling and branch blocking do not result in new endpoints Depending upon how you reorder you may also be able to reorder nodes in an asymmetric tree and play the endpoints Therefore you can often make significant changes to your model and still be able to play a recorded Endpoint Database This section demonstrates how you can play endpoints to s
183. anded from is still selected the Expand button will no longer be enabled in the ribbon or the context menu The Collapse button is now enabled on the ribbon There are two collapse options Collapse and Collapse and Zoom Analogous to Expand the Collapse command will collapse the tree to the previous level This can also be accomplished by double clicking the node While you have expanded above there are still deeper levels that are not shown Select the Marketing Costs node in the top right corner Click Policy Display Expand Subtree gt Click the Zoom Full button on the status bar located at the lower right corner of the screen The Policy Tree should now look like Figure 3 4 Now the Policy Tree is expanded so that we can see nine endpoints displayed as blue triangles at the top right corner of the tree This model contains 108 endpoints or paths so the tree would be difficult to read if 80 Syncopation Software we displayed them all at once Practical applications often involve thousands or millions of endpoints so it s important to be able to manipulate the Policy Tree to show the desired information Marketing Costs Nominal 24 4 Cost per Unit Low 118 3 i 30 Unit Sales In house 107 7 Product Strategy 107 7 Marketing Costs High 9 2 40 Marketing Costs Nominal 119 9 Cost per Unit Nominal 107 7 60 Marketing Costs a High 89 6 40 Marketing Cost Nominal 21
184. appens about 70 million is lost by not launching the product anyway The overall loss in expected value is 15 x 70 million or about 11 million Market Share Launch Yes 394 9 Failure 394 Ar 122 9 30 No 174 lig Key competitor outcome Markot St j Aarket Shar FDA Approval _Yes 195 8 e ARRA T gt 3 09 50 Launch _ Yes 325 7 Develop Yes 10 9 W i 494 y 7 Disappointment 323 7 t 10 9 174 0 40 No 174 0 lt 1 i P Breakthrough 174 0 i A lt Ae P No 174 0 i i 50 lt 0 0 lt Figure 6 35 Policy Tree for Revised Model 213 Chapter 6 Asymmetric Trees Syncopation Software 214 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models 7 CONDITIONING AND LEARNING IN DECISION MODELS When a decision is preceded by a chance node that provides information we say that the model contains learning For example before deciding whether to carry an umbrella you may look out the window and see if there are clouds Learning about the clouds uncertainty provides imperfect information about the uncertainty of concern whether or not it will rain later that day Other typical examples of learning in decision models include medical tests market research and sampling of the physical environment Test drilling for oil is a classic example of learning and is the basis for the tutorial in this ch
185. apter This chapter assumes you have already been through at least one of the earlier tutorials Chapters 2 and 3 or Chapter 5 in this manual or that you are already familiar with DPL Therefore basic procedures such as starting DPL and opening and saving workspaces will not be explained in detail here This chapter also assumes some familiarity with basic probability theory and Bayes Rule The example in this chapter demonstrates DPL s capability to perform Bayesian revision which can be simply described as reversing the order of conditioning among chance nodes to take advantage of the data available Please consult online Help and or a probability theory or decision analysis textbook if you need further explanation of these topics 215 Chapter 7 Conditioning and Learning in Decision Models Syncopation Software 7 1 INCORPORATING IMPERFECT INFORMATION IN A MODEL In this section you will enhance a simple decision model by adding a node that represents imperfect information about an important value driver You will see how the learning associated with such nodes can significantly change the measured value of an asset Start DPL gt Open the workspace Learning da It should be in the Examples folder underneath where DPL is installed C Program Files x86 Syncopation DPL8 Examples if the default installation location was used See Figure 7 1 r id DPL Learrung Wildcat omple O amon
186. apter 3 Decision Analysis Results 3 4 EXPECTED VALUE OF PERFECT INFORMATION CONTROL The expected value of perfect information control is actually two outputs which both indicate how important it is to reduce the uncertainty associated with each discrete chance node in a model The expected value of perfect information indicates how valuable it would be to have perfect information for a chance node before making a decision The expected value of control indicates how valuable it would be to have complete control over the uncertainty You will now examine the expected value of perfect information control results Double click on the Value of Info Control item in the Workspace Manager to activate the window The Value of Information Control output as shown in Figure 3 27 is displayed Expected Value Sales 2 Sales 1 Price Cost per Unit Marketing Costs E Information Control Figure 3 27 Expected Value of Perfect Information Control 107 Chapter 3 Decision Analysis Results Syncopation Software DPL displays the value of perfect information for each chance node as a red vertical bar and the value of control as a yellow vertical bar By default both sets of bars are displayed and the nodes are sorted in descending order by their value of perfect information DPL calculates the expected value of perfect information for each chance node by moving it to the front of the decision tree or as far t
187. ar data set in which the records or rows correspond to endpoints single paths in the Decision Tree and the fields or columns correspond to information about each endpoint such as the state of each event for that path the values of the objective function and each attribute for that path and so forth With DPL Professional you can record view sort and re use an Endpoint Database Endpoint Databases can also be imported from or exported to other applications An Endpoint Database can be re used for further analysis this is called playing endpoints In order to play endpoints you must select Full Tree Enumeration from Endpoints for your evaluation method In a number of situations even if you make changes to your model after recording endpoints you can still play the endpoints to produce additional results With large DPL models and or DPL models linked to large spreadsheets the ability to record and play endpoints can save you a lot of analysis time The first two sections in this chapter provide an overview of the Endpoint Database and its display report and demonstrate how to play recorded endpoints The last three sections describe how to reconnect export and import an Endpoint Database A model with a large number of endpoints will result in large Endpoint Databases The displaying of a sizeable Endpoint Database in a report at the completion of a run consumes additional memory and may take some time In addi
188. ar with the spreadsheet 144 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models Al A T e i D E F G 5 an a al Y p Enlightify Consulting LLC Actual Actual Fereesse 1013 1004 Told 301s m 1017 2018 2019 2070 yiil Cen DCF Calculation tm im im 4m im m im im im Free Cash Flow Revarnwasr Trendire Revenue Tirade R ve ige Growth Rota Adjustment for bad debe Bad Dither Alewanesr A af Reet Adjudiment far rev regulation Share of Reverse at Fine due ce New Repos Shore of of Rak Revenue Low Tol Revenues Working capital interest Werking Cepte jays fntevea Rote ani Working opra Controllable oora Deher Gear Groat Aor Allocited cr rh rds Ocher Cen Greath Aone Litigation Tool Comm Met Cash Flew Not Present Value NPY Discount Rate Present Value Farmor Protem Waluo of Cash Flown Figure 5 4 Cash Flow Model DCF Calculations 5 2 CREATING LINKED VALUES gt Switch back to DPL To begin building a DPL model linked to the Enlightify spreadsheet you ll first need to create some linked values gt Click the Model Links Add icon Note that the default spreadsheet link type indicated by the icon should be Excel Calculation Link If it isn t use the split button drop down to select it from the list gt In the Create Values from Excel dialog click the Browse button and browse to find Enlightify1 xls See Figure 5 5 Clic
189. arge complex models A few conventions have been used in the text that follows An instruction to you in a tutorial will be contained in a bulleted paragraph Information to be entered in edit boxes Excel cells etc is contained within double quotes Do not include the double quotes when entering the information DPL ribbon tab names groups and commands are separated by vertical bars For example Model Node Add refers to the default command indicated by the icon in the Add split button located in the Node group on the Model tab of the command ribbon When a command other than the default indicated in the icon is desired you will be prompted to drop down the split button and select the command from the drop down list Chapter 1 Introduction Syncopation Software The word Options following a vertical bar refers to the dialog box launcher which is located at the bottom right corner of the given tab group For example Model Node Options refers to the dialog box launcher for the Node group on the Model tab of the ribbon and will launch the Model Options Dialog If there are multiple tabs within a dialog then the item following the last refers to the desired tab For example Model Links Options Settings refers the Settings tab of the Model Links dialog 1 2 HOW TO GET HELP WITH DPL Online help contains detailed descriptions of DPL features as well as a reference section for DPL dialogs Descripti
190. art item in the Workspace Manager labeled Expected Value 4f 337 Chapter 12 Advanced Decision Analysis Results Syncopation Software DPL activates the Risk Profile chart for the Wildcat model as shown in Figure 12 15 The Risk Profile Chart indicates that in about 10 of the scenarios the objective function exceeds 200 gt 2 a a v 8 O 0 100 80 60 40 20 0 20 40 60 80 100 120 140 160 180 200 220 240 260 Figure 12 15 Risk Profile Chart for Wildcat Model You will now use DPL s Filter by Objective Function feature to find out more about which scenarios lead to big gains Double click the Policy Tree item in the Workspace Manager to activate it again gt Select Policy Policy Summary Compare By Objective Function The Filter by Objective dialog appears as shown in Figure 12 16 358 Syncopation Software Chapter 12 Advanced Decision Analysis Results Filter by Objective Function The baseline for comparison is the set of endpoints that are included in the optimal decision policy Filter the endpoints by selecting those endpoints that are included in the optimal decision policy SAND for which the objective function falls within the range below Leave Min or Mas blank to not restrict the minimum or masimu Mlan Full range 291 6875 Filter range 121 75 291 6875 Cancel Figure 12 16 Filter by Objective dialog DPL displays the full range of the
191. as needed and then proceed to Chapter 1 Time Series Percentiles 253 Chapter 9 Multidimensional Value Node Syncopation Software 9 1 CREATING AND LINKING A ONE DIMENSIONALVALUE NODE This section contains brief tutorials on setting up one dimensional Excel linked array and series value nodes in DPL One dimensional Excel linked arrays and series can be either import or export nodes The tutorial will look at an example of a one dimensional array import node a series import node and a series export node 9 1 1 A One Dimensional Import Node You will begin by examining a model of a decision about a gas pipeline investment Select File Open gt Navigate to the Examples folder where you installed DPL The default location path is C Program Files x86 Syncopation DPL8 Examples Select Multidimensional Value da and click Open You should see the Energy Model as depicted in Figure 9 1 254 Syncopation Software Chapter 9 Multidimensional Value Nodes f W W DPL Multidimensional Value Energy Model EE Professienal Version _ Energy Model x 4b Fot Help press Fl Zoom 65 full Prev L Figure 9 1 Multidimensional Value Model This model contains a few features that may be unfamiliar to you depending on your experience with DPL You don t need to understand all the logic in the DPL model or the linked Excel cash flow model in order to complete thi
192. ases make a decision make a max decision or make a min decision The first three are equivalent You can use the one that makes the program easiest to read Next you identify each alternative or group of alternatives with the keywords on to or about You can use whichever keyword makes the program easiest to read and you can use different keywords with different alternatives of the same decision decide to Release Product es to Release Product No 412 Syncopation Software Chapter 15 DPL Programs and Code 15 5 2 Chance Events You specify a chance event in the sequence section of a program with the keywords chance or gamble or the keyword phrases take a chance or take a gamble All of these keywords or keyword phrases are equivalent You can use the one that makes the program easiest to read Next you identify each outcome or group of outcomes with the keywords on Or of You can use whichever keyword makes the program easiest to read and you can use different keywords with different outcomes of the Same chance event gamble on Market Forecast High pss on Market Forecast Medium soss on Market Forecast Low 15 5 3 Branches If a decision or chance node is symmetric DPL code provides a way to use schematic notation as in a Decision Tree The event name appears by itself and the individual states are not listed For example gamble on Demand then The above code is equivalent to the node being pla
193. ashed line will appear to preview the bended arc gt Continue dragging until your arc looks something like the arc in Figure 2 21 Release the mouse to fix the arc in place Product Strategy In house NPV License A NPV v Figure 2 21 Model with Bent Arc When you defined conditioning for Revenue per Unit you used the Conditioning dialog DPL also provides a way for you to define conditioning by changing the arcs in the Influence Diagram You will now use this procedure to condition Cost per Unit and Marketing Costs on Product Strategy 46 Syncopation Software Chapter 2 Building Decision Analysis Models Click Model Influence Arc Add on the ribbon see Table 1 3 Draw an arc from Product Strategy to Cost per Unit click on Product Strategy and then click on Cost per Unit cy Double click the new influence arc The Value Influence Type dialog appears cy Check the Different values for each conditioning event state checkbox as shown in Figure 2 22 Value Influence Type EJ Values est values for each conditioning event state Make this the default value influence type Cancel Figure 2 22 Value Influence Type Dialog gt Click OK to close the Value Influence Type dialog You have now told DPL that Cost per Unit is conditioned by Product Strategy Double click Cost per Unit The Node Definition dialog appears with the Data tab selected Not
194. ation DPL has suggested using the Nominal state for both the Low and Nominal tornado diagram setting Leave this setting as is Note to start DPL chooses the outcome defined to be Default in the Node Definition dialog to be Nominal for a Base Case Tornado the first outcome for Low and the last outcome for High For discrete chance nodes with more than three outcomes you may need to change the settings in the Base Case Tornado Setup dialog gt Click OK to run the Base Case Tornado Diagram DPL produces the Base Case Tornado Diagram as shown in Figure 2 38 In this example the Base Case Tornado is similar to the Value Tornado that you ran on the deterministic model i e the model with no discrete chance nodes in Section 2 5 This will not always be the case for several reasons First the Base Result of a Value Tornado may not be the same as the Base Case for the Base Case Tornado if the states chosen as Nominal for the Base Case Tornado do not match what was used as the current values in the Value Tornado Second if you change the high or low values for value nodes that you have changed to discrete chance nodes the two tornado types will differ 68 Syncopation Software Chapter 2 Building Decision Analysis Models Base Case Tornado Unit Sales Marketing Costs Cost per Unit 60 80 100 120 140 160 EE Low Policy Change E Policy same as Nominal E High Policy Change Figure 2 38 Base Case Tornado Diagram Base
195. ave seen the Endpoint Database report looks a lot like an Excel spreadsheet In fact you can quickly and easily export the Endpoint Database to a comma separated value CSV file which can be read by Excel or other applications 330 Syncopation Software Chapter 11 Using the Endpoint Database There are two ways to export endpoints with state names for each event as you see in the report or with state numbers If you plan to use the CSV file in Excel or another application where you need to see the state names you can use the state names option However if you wish to import the CSV file back into DPL later as you will do in the next section you need to choose the state numbers option You will do both in this tutorial gt If you have the energy model open with a recorded Endpoint Database from the previous sections double click on the Endpoint Database item to make sure the report is displayed Click Yes if you are prompted to display the report gt If you do not have the energy model currently open or you do not have a recorded Endpoint Database open it and or run it and record endpoints as described in the beginning of this chapter Activate the Endpoint Database if it is not already active gt Go to the Data tab drop down the Export list and select Data Choose a folder location and name the file Endpointsi csv Click Save You will be prompted whether to use state names or state
196. ave time when you need to do what if analyses or quickly test changes to your model Before beginning this section you should have an Energy Model Endpoint Database recorded and available from the previous section 315 Chapter 11 Using the Endpoint Database Syncopation Software 11 2 1 Reordering Nodes You will reorder the tree so that Operating Costs are known before the Pipeline Strategy must be decided gt Switch back to the Model window Ctrl F12 or double click on the Energy Model item in the Workspace Manager Click on the Operating Costs chance node in the Decision Tree cy gt Drag the node so that the chance node is positioned over the Pipeline Strategy decision node cy Release the mouse button The reordered tree should look like Figure 11 8 Note you can also reorder by right clicking on a node and choosing Reorder Pipeline Strategy Market 2 Operating Low Appraisal Forecast Sani Costs Mkt 1 10 BCM Total_NPV Results Price say al 4 oe Nominal Nominal Nominal TOCEE Low Mkt 1 20 BCM Total_NPV I High High High Stable Yes Nominal ig g g Nominal Nominal Total_NPV Market 2 High Unstable j Political High High Mkt 1 and 2 10 BCM Mkt 1 and 2 20 BCM Mkt 1 and 2 30 BCM Upfront_NPV Figure 11 8 Reordered Decision Tree Inthe Home Run group select Risk Profile and Policy Tree Drop down the Home Run Decision Analysis split button and select Full Tree Enumerat
197. aving changes it is not needed for the last section of this chapter 299 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software 10 3 USING MULTIDIMENSIONAL VALUE NODES IN GET PAY EXPRESSIONS As discussed in Chapter 8 and earlier in this chapter any time you use DPL s multiple attributes feature whether to set up Time Series Percentiles or to track multiple attributes for the objective function you must set up your Get Pay expressions to match the attributes you have defined On any decision tree branch that has a get pay you need to have exactly as many Get Pay expressions as the number of attributes you have defined If you use multidimensional value nodes arrays or series in a multiple attribute model you may find that you need to refer to elements of these value nodes in your Get Pay expressions This section provides a brief discussion of how to do this using the Get Pay tab of the Branch Definition dialog and its Fill Down diah button which you also used earlier to set up Time Series Percentiles Assume that you have set up a project evaluation model in DPL with six attributes The first attribute corresponds to the objective function and is called NPV Management is concerned about the up front spend on this project so you must also track the other five attributes which represent corporate investments in each of the first five years of the project You have left the attribute names at their de
198. b DPL will warn you that default probabilities have been used when you first run the model and will ask whether you wish to continue the run with those probabilities As you Saw earlier you can select multiple nodes in the Influence Diagram When changing node types you can select multiple unconditioned value nodes and change them to discrete chance nodes or vice versa In this example we could have selected both Unit Sales and Price and changed them simultaneously to discrete chance nodes If you d like to try this now select them both hold down Ctrl while selecting change them to value nodes Model Node Change To Value You will get a warning Click OK Then change them back to discrete chance nodes Model Node Change To Discrete Chance If you try this and don t review the node data you will get the default probabilities warning mentioned above Your model should now look something like Figure 2 36 DPL has updated the Default Tree further by adding the discrete chance nodes for Cost per Unit Marketing Costs and Price and by moving the NPV get pay expression to the end of the tree on the branches for Price 65 Chapter 2 Building Decision Analysis Models Syncopation Software Marketing Product Costs Strategy i Price Low Low In house NPV Nominal Nominal License High Figure 2 36 Model with Three Discrete Chance Nodes You are now almost ready to run the modified model with uncertainty inc
199. ble 16 3 Unsupported Functions Conversion Options definitions If you change any of the options your new settings will be used the next time you convert unless you change them again You can click the Defaults button to revert to the default settings Leave the settings as they are gt Click Cancel gt Click OK in the Convert Spreadsheet to Program dialog DPL creates a DPL program file with the spreadsheet converted to DPL code in it For large spreadsheets this may take some time While converting DPL displays the Converting dialog which allows you to cancel the process The spreadsheet you just converted probably converted too quickly for you to even see the dialog The DPL program with the converted spreadsheet code has the same name as the spreadsheet i e Cash Flow1 xls in this instance After conversion DPL removes the spreadsheet from the Linked Spreadsheets list box and puts the name of the DPL program with the converted spreadsheet code in the DPL Calculation Program edit box gt Click OK to close the Model Links dialog Note that DPL has created an item in the Workspace Manager for the DPL program with the converted spreadsheet code DPL places the item for the DPL program under the model that references it See Figure 16 5 The icon for a referenced program has a link on it indicating that it is linked to the model beneath which it is displayed 431 Chapter 16 Converting Spreadsheets to DPL Code Syncopat
200. bo Box Only enabled for multiple attribute models Use the drop down list to select a single attribute for the objective function or type in an expression Max Select attribute W f involving more than one attribute and or value from the model el vi Use Max Min drop down list to Objective amp Utility either maximize or minimize the BONET v The dialog box launcher will open the objective function Use the W Objective tab of the Objective amp ull h or buttons to insert a Utility dialog where you can add variable or function rename and delete attributes and specify an objective function and or constraint function You can also i choose to display Fat Policy within Built In Risk Tolerance Combo Policy Tree Within the Risk Box Tolerance tab you can specify a Specify the coefficient for the single risk tolerance for the entire built in risk tolerance function in tree order to model attitudes towards risk Use the h or 4 buttons to insert a variable or function Table 1 7 Model Objective amp Utility Ribbon Group 18 Syncopation Software Chapter 1 Introduction View Tab You can access several useful display commands within the View tab Once you have multiple items within your workspace you can arrange windows and icons and switch between windows via commands within the View Window group Within the View Layout group there are commands to Snap nodes to the grid and a
201. ccseeseseeesesneenenes 189 calculation links establishing cccceeeeeeeseeeeeeees 438 case SENSILIVILY cceceeeeeeeeeeeeeeees 394 certain equivalent seseseseseserenen 370 chance keyword ssssssssersrsrerennn 400 chance node CODUMUOUS saeisntenarevedareasyiuuiins 159 discrete eee 59 change node types See node types changing Command WINCOW ccsceeeeeeeeees 424 COMMENUS saciinmnmiiavariacterenierrctant 393 COMPANG soscontransncdnencomnrers See Policy Summary Comparison compatibility cceeseeseeeeeseee eens 455 Conditioning via arc type sssssssnrssurnnrrraninnnnnna 46 via Node Definition 4i const keyword ssssssssssserersrerennn 410 constraint function s s sesssssesenn 250 continuous chance node 159 controlled KeyWOrd ssssecseeeeees 403 Conversion OptiOns ccccseeeeeeees 428 converting Models to Programs 00008 394 Spreadsheets to Programs 425 data input tree selection IM cesececeseceeeneeereeenens 42 decide KEYWOMK ccsccecseeeeeeeeeees 412 decision keyword scceceeeeeeeeeeeees 400 decision node ceceseceeeseeeeenenenees 37 479 Index decision sensitive ceceeeeeeeeeeeeees 58 decision tree attaching nodes ccsceeeeeeees 189 MOVING NOCES ccceeeeeeeeeeeeeees 189 Decision Tree pane ccceeeeeeeeee ees 8 default arc tyPeS cccccseeseeseeeeees 222 default State ccscssec
202. ced in the Decision Tree as shown in Figure 15 1 Demand Low Nominal A High Figure 15 1 Node Placed with Symmetric Grouping 413 Chapter 15 DPL Programs and Code Syncopation Software If a decision or chance node is asymmetric then each state must be listed gamble on Demand High then on Demand Medium then on Demand Low then The above code is equivalent to the node being placed in the Decision Tree as shown in Figure 15 2 Demand Nominal Figure 15 2 Node Placed with Asymmetric Grouping In the asymmetric case DPL will generate an error message if you do not describe every alternative associated with a decision or chance event If the name of a decision or chance event appears with the name of only one state the names are separated by a period If the name appears with the names of two or more states the name is followed by a period and a list of the state names The state names are enclosed in braces and separated by commas gamble on Demand Low Medium then on Demand High then The above code is equivalent to the node being placed in the Decision Tree as shown in Figure 15 3 414 Syncopation Software Chapter 15 DPL Programs and Code Demand Nominal lt Figure 15 3 Node Placed with Mixed Grouping The keywords representing a decision s alternatives or a chance event s outcomes must start in the same column Also they cannot start in the same column a
203. cel aiiin Home Insert Page Layout Formulas Data Review View DPL A B o ep 8 l oe e Ss8F oa iene EAP GA a7 B Z U AV 2323 Bir g Delete 8 7 7 Paste g z re ear Styles Sort amp Find amp F iy Ar A te v 00 0 Formate Q gt Fitter Select Clipboard amp Font x Alignment P Number P Cells Editing Color coded blue yellow or green gt currently linked to DPL Basic Parameters Discount Rate Market 1 Growth Model Choice Market 2 Growth Model Choice Market 1 Size Year 1 Market 2 Size Year 1 Annual Growth Market 1 Option 2 Annual Growth Market 2 Option 2 Ill Annual Growth Assumptions for Market Growth Models 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 Market 1 Size Option 2 110 116 121 127 134 140 147 155 163 171 Market 2 Size Option 2 180 189 198 208 219 230 241 253 266 279 0 04 005 005 005 004 004 003 0 03 0 03 0 03 Year by Year Growth for Market 1 Option 3 0 04 0 05 0 05 005 004 004 003 0 03 0 03 0 03 Year by Year Growth for Market 2 Option 3 Other Assumptions Annual Price Growth Market 1 i 2 O Annual Price Growth Market 2 OO O O O Annual Op Costs Growth Market1 4 o Annual Op Costs Growth Market2 3 Start year OOo 203 O value p O 055 value q O O Upfront Cost Escalation gt Ha Initial Marketing Spend i e 35 4 4 gt gt Assumptions DCF Market Trend Data 2 i
204. cision Sensitive A variable in a model such that when its value is changed a different decision policy is optimal from that which is optimal for the variable at its current setting Decision Tree A graphical representation of a decision model that displays the sequence of events including the order of decisions and uncertainties and when get pay expressions occur Decision Tree pane A pane of the Model Window that provides a graphical interface to manipulate the decision tree in a model Default State The state of an event that is user defined as the default setting for purposes of linking to an Excel spreadsheet If the state of an event is unknown for a particular path on an asymmetric tree DPL assumes the event is in its default state Default Tree The decision tree that DPL builds automatically in the Decision Tree as you develop a model in the Influence Diagram DPL builds the Default Tree based on the decision nodes chance nodes value nodes and influence arcs you define in the Influence Diagram The Default Tree is always symmetric DPL will continue to build the Default Tree until you edit the decision tree yourself directly in the Decision Tree Definition Section A section of a DPL program which specifies and initializes the elements of a decision model decisions uncertainties variables values series arrays etc The Influence Diagram is a graphical representation of a definition section Dependency Node B depen
205. cribed in Section 5 7 182 Syncopation Software Chapter 6 Asymmetric Trees 6 ASYMMETRIC TREES This chapter will walk you through the process of building an asymmetric decision tree Asymmetric decision trees are important because many business decisions are asymmetric in nature different strategic alternatives are often influenced by different sources of uncertainty Modeling that asymmetry directly usually improves computational efficiency and results in a model which reveals the structure of the problem and is useful for communication This chapter assumes you have already been through at least one of the earlier tutorials Chapters 2 and 3 or Chapter 5 in this manual or that you are already familiar with DPL Therefore basic procedures such as starting DPL and opening and saving workspaces will not be explained in detail here Note that in the tutorials in Chapters 2 3 and 5 you worked mainly in the Influence Diagram pane of the Model Window In this chapter you will be working mainly in the Decision Tree pane because asymmetries are modeled in the decision tree Refer to Chapter 1 if you need a refresher on how to move around the Model Window 6 1 ELIMINATING REDUNDANT PATHS You will begin with a simple symmetric model for a pharmaceutical R amp D project Start DPL Open the file AsymmetricTrees da If the default installation folder was used when installing DPL then the file should be in the C Program Files
206. ct Land_Weight from the Value for sensitivity drop down list Click OK Type 3 0 for From Type 0 0 for To Type 0 25 for Step size The Run Rainbow Diagram dialog should look like Figure 8 12 GUC 4 246 Syncopation Software Chapter 8 Objective Function and Multiple Attributes Rainbow Diagram Wildcat Vanable Varable name Land weight Curent value 0 5 From 3 0 O Number of steps 13 Step size 2 Evaluation method Fast sequence evaluation Figure 8 12 Run Rainbow Diagram Dialog gt Click OK 247 Chapter 8 Objective Functions and Multiple Attributes Syncopation Software The Rainbow Diagram Figure 8 13 shows that Land Weight would have to be less than 2 for the decision policy to change This means the policy of testing before deciding to drill is fairly robust there would have to be a substantial change in the attribute weighting for it to change N N N w 2 gt ge D a U D Q x i m co H a j A mens N QW 1 75 1 5 Land Weight Figure 8 13 Rainbow Diagram on Land Weight When you change Get Pay expressions in the Decision Tree by default DPL expands the branch length of the node where the Get Pay expression appears You may not have noticed this in the model you just modified because the Get Pay expressions were not particularly lengthy If you do have lengthy Get Pay expressions you may end up with a tree tha
207. cusses how to do Monte Carlo simulation in DPL gt If needed start DPL DPL will start with a blank Workspace See Figure 5 1 oden o 8 ids OPL Workspace Modeli om f_ 4 V we ae f Detour Tree aa gt Stighten nitance nioi all Mt act l sa Eat Chanoe is Ada Frem to treat Unzontrol All Workspace Manager a A Weekipece gt Modai Perpetual For Heip press Fl Figure 5 1 A Blank DPL Workspace 141 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software The window on the right hand side of the screen is the Model window The left hand pane is the Workspace Window For more information about the DPL Workspace Window see Chapter 1 At the moment Model1 is the only document in your Workspace The risk analysis model you will build will consist of a DPL Model used to characterize uncertainties and an Excel model used to calculate cash flows 5 1 THE EXCEL CASH FLOW MODEL Before creating the DPL model take a brief tour of the Excel model provided for this tutorial gt Start Excel as you normally do Open the file Enlightifyi xls This file will normally be found in the Examples folder where DPL was installed e g C Program Files x86 Syncopation DPL8 Examples if the default location was used during installation 142 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models Enlightif
208. d Decision Analysis Results Gud DPE Wildcat da wildcat o EJ Workspace Manager 7 Workspace gt wildcat OPL Decana Progra a _D coe A S D DSA S DE J fama 1 ante tT a Se j omn LerToe 0 zs vse im DORDE paea Iss at est F Figure 12 11 Wildcat Model The Wildcat model is a simple model of oil exploration The decision tree is asymmetric see Chapter 6 There is an upfront decision to test one of two versions or not If a Test is conducted information on the seismic structure is gained which provides an indication of the amount of oil If no test is conducted this information is not gained After either testing or not uncertainties regarding oil price and pipeline costs are resolved These are followed by a decision to drill or not and further uncertainties regarding drilling costs and the amount of oil The results of the two tests are conditioned on the actual amount of oil present however the results of the tests are available before the amount of oil is known DPL uses Bayesian revision See Chapter 7 to calculate the unconditional probabilities of the test outcomes gt In the Home Run group check Risk Profile and Policy Tree only Make sure to uncheck Init Dec Alts gt Click Home Run Decision Analysis Fast sequence evaluation should be indicated as the default 3353 Chapter 12 Advanced Decision Analysis Results Syncopa
209. d as indicated in Figure 4 15 a Node Definition Strategy Commands Print Page Setup Copy Format Full Screen General Feature SetA F Status Quo P Feature Led Add features P Price Cuts Keep same Differentiate Product Minimal Feature SetB Add features Keep same Minimal Feature SetC Add features Keep same Minimal Marketing Spend Increase AMO Keep Same Decrease Price Change A Increase AMO Keep same Decrease Increase AHO Keep same Decrease Price Change Cc Increase BRO Keep same Decrease Figure 4 15 Status Quo Strategy Defined by Decision Alternatives 136 Syncopation Software Chapter 4 Strategy Tables Define the remaining three strategies as indicated in Table 4 1 In the table below Add Add Features Same Keep same Min Minimal Inc Increase and Dec Decrease Feat Feat Feat Market Price Price Price ure ure ure ing An oe ue Set A SetB SetC Spend Differentiate Min Same Add Inc Products Table 4 1 Strategy Definitions After completing the strategy definitions your Strategy Window should look like Figure 4 16 Note that each strategy has to be defined by an alternative of each decision in the strategy table i e the icon representing the strategy has to appear in each decision column However not every decision alternative needs to b
210. d maximum of the range bar 293 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software 0 J 10th 90th Percentile 95th Percentile 50th Percentile 10 Sth Percentile Expected Value T T T T T T T T T T 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 Figure 10 15 Time Series Percentiles Chart with Range Bar The Time Series Percentiles Chart does not provide all the information you might need about the full risk profile for each of the ten attributes cash flows it gives you up to seven percentiles plus the expected values However more information about the uncertainty in each attribute is available by generating Risk Profiles from a Decision Analysis run You will examine this feature in the next section 294 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics 10 2 RISK PROFILES WITH MULTIPLE ATTRIBUTES In the previous section you set up a Time Series Percentiles analysis to look at ten attributes years of cash flow in the energy model In some situations you might want more detailed information about the full risk profile range of uncertainty for one or more attributes You will continue the example from the previous section to see how easily this can be done with one run of DPL Activate the Energy Model in the Model Window gt In the Home Run group make sure that Risk Profile and Policy Tree are the only selected out
211. defining them in terms of the decision alternatives that each represents To define strategies in terms of a decision alternative you drag the strategy icon to below the decision alternative that defines the strategy for each decision in the strategy table You can either drag the strategy icons from the strategy column on the left to the decision alternative columns or within the decision columns from one alternative to the other Drag the red square icon representing Status Quo from either next to Status Quo or under Feature Set A Add features to below the Keep same alternative under Feature Set A See Figure 4 14 135 Chapter 4 Strategy Tables Syncopation Software Node Definition Strategy General Print Page Setup Copy Format Full Screen Commands Price Change Cc Price Change B Price Feature Change A SetC Marketing Spend Feature SetB Feature SetA BES Pl Price Cuts Differentiate Products B Minimal Increase DARO Add features Increase Increase Increase EANO AANO HERO Hae Add features PARAO Keep same Keep same Keep Same Keep same Keep same Keep same Minimal Minimal Decrease Decrease Decrease Decrease Figure 4 14 Dragging Strategy Icons to Define Strategies Repeat this procedure to drag the icon for Status Quo to the Keep same alternative of the remaining decisions The Status Quo strategy should now be define
212. del with 1 x 10 Value Node Note when linking a one dimensional value node to a named range in Excel you may use either an array or a series node DPL automatically creates a series node when it finds formulas in the range But if you were creating the node and linking it using the Node Definition dialog you can use either type to link the node to Excel See Figure 9 9 for an example of what the Node Definition Data tab looks like for a one dimensional import array 262 Syncopation Software Chapter 9 Multidimensional Value Nodes Node Definition Ten Year Cash Flow General Links Columns 10 E Seres interval entry O Create Mx 1 column vector fta v fi Colurnn Expression a 1 3 3 4 5 6 T i g K J Cac He Figure 9 9 Data Tab for One Dimensional Import Node As your model is now set up the Ten Year Cash Flow value node will import ten years of cash flow values from the Excel spreadsheet each time the model is run However further set up of the model is needed before you see these values in analysis results At this point you might want to analyze the expected values and ranges of uncertainty around these annual cash flows To do this you need to add 10 attributes to your model to correspond to the annual cash flows Defining attributes in your model allows you to run Time Series Percentiles to visualize how the uncertainty in the cash flows evolves over time
213. ding and Analyzing Monte Carlo Simulation Models 5 6 MODELING AN UP FRONT DECISION Thus far your analysis has focused on the risks to the business without considering any decisions management might be able to make to mitigate those risks Enlightify s parent organization has considered spinning off Enlightify Doing so would alleviate regulatory pressure but would probably reduce its rate of growth because cross selling opportunities synergy would be lost Next you ll add this decision to your model Before making this change you want to make a copy of the model as it is gt In the Workspace Manager window right click on Risk analysis and select Duplicate gt Rename the new model Decision Analysis by right clicking on it and selecting Rename If necessary drag the splitter up from the bottom of the Influence Diagram to make the decision tree pane visible Your Workspace should look like Figure 5 25 ee eS i O SS ad oe DPL Dn ghtiy da Decision Analysis EJ Add to i Deris Eh selects z F Rink Prafile Policy Tree Levels inene Meme Carts tampir T ndi Tne Ze D 44 topy gt Prei rplice _ Ghedi hain Policy Sumiaery pripili Palin 1U 7 Ci mp r Decises Bankban le Gatien y Eu Pate ee Arik rit Dec Ata Pia T Vaip Comelabens Restart 100 T hm 5 z Workspace Mani m ad I Enkegeet da i go Beberranistic ee Pde nets sa PN Sme POOO Sarrik izpiitid Value ROH
214. ditioning event changes to its second state After associating numbers with all the states of the last conditioning event DPL varies the state of the next to last conditioning event and so on This continues until a number is associated with all combinations of states Once again the comments are ignored by DPL and do not determine the order in which you specify the values 403 Chapter 15 DPL Programs and Code Syncopation Software 15 4 8 Shortcuts Statenames If you don t want to name the states of an event you can define events without state names DPL will supply default state names For example the following statement defines a decision called Invest that has two alternatives decision Invest 2 In the program you can refer to these alternatives as Invest s1 and Invest s2 If an event has only two states DPL also allows you to call them t for true and f for false or y for yes and n for no respectively If you do not specify the number of alternatives or list their names DPL will generate two unnamed states The following definition is equivalent to the one above decision Invest If you define a chance event without specifying state names you should still provide probabilities chance MarketPrice 4 0 1 0 3 0 4 If you define a chance event without probabilities DPL will assign a uniform distribution to it chance MarketPrice 4 For this example each state of MarketPrice will have a probabi
215. ds on node A if one of the data expressions for B includes A but B does not have a separate data expression for each state of A For example if Revenues Units Price then Revenues depends on Units and Price See also Conditioned 463 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software Deterministic Model A model that only has deterministic relationships there are no uncertainties The model contains only value and decision variables and yields a single output for a single set of inputs or setting of each variable A spreadsheet is an example of a deterministic model Discrete Chance Node A node in the Influence Diagram or decision tree that represents an event whose outcomes are uncertain A discrete chance node models the uncertainty of the event with a discrete number of states or outcomes Probabilities are specified for each state or a named distribution is given The graphical symbols for a discrete chance node are bright green ovals in the Influence Diagram and bright green circles in the Decision Tree Discrete Tree Simulation Simulation method of evaluating models utilizing random sampling of discrete or discretized probability distributions chance nodes with a finite number of outcomes See also Monte Carlo Display Function A function that enables you to write text or formatted numbers to the Session Log during an analysis allowing the writing of custom reports Distributed Sampling A
216. e Adding Learning Events You can add a chance event to the model and play endpoints as long as the new chance event only probabilistically conditions or is probabilistically conditioned by other chance events in the model If the conditioning is value wise you will not be able to play endpoints In the Influence Diagram pane add a chance node called Op Costs Research and position it near the Operating Costs chance node On the General tab specify two outcomes Low Costs and High Costs Set Low Costs to be default gt Switch to the Data tab gt Click the Conditioning button Select Operating Costs Click OK Edit the probability data for Op Costs Research so that it looks like Figure 11 10 ere fee Ops e Research General Conditioning Print Page Setup Copy FullScreen Wheel iultipl qla QA Distribution type discrete v Links Probability J vif Value vi Ops Costs Low Costs Low Research 9 High Costs Operating Ops Costs Low Costs Costs Nominal Research 5 High Costs 5 Ops Costs Low Costs High Research a High Costs Figure 11 10 Probability Data for Op Costs Research Node Click OK The new node appears in the Influence Diagram and Decision Tree as shown in Figure 11 11 319 Chapter 11 Using the Endpoint Database Syncopation Software Figure 11 11 Influence Diagram and Decision Tree with Op Costs Research Node Notice that the new
217. e l l l l b b l b l l l l l Create arrays based on range sIZe Prefix node name with sheet name Figure 5 6 Named Cells in Enlightify1 xls 146 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models gt Click OK DPL will create a value node for each of the selected cells A value node is a blue rounded rectangle representing a constant or calculated quantity See Figure 5 7 m S Sid DPL Werkspacel Enlightity a Fie Add to WS Delete O Sete a s Ji Riik Profie Ji Ponc Tite Levels nome Mome Cario Samples Rename a Copy W FinaRepiace _ Objectwe Function Policy Summary Endpoints viel 1000 hd Compile Decitran Weekspace Manager 3 S Workipace _ Unlightity1 Session Log OFL Decana Prograrenng Lacgeage Copyright 2003 2013 by Syocopstion t ncerpersiing sahware on bcanse to lle Syncopation Software inc le fs m I Fee Help pios A goo WE Fuil Prev Figure 5 7 Value Nodes Created from Enlightify1 xls Named Cells The top pane of the Model Window is called the Influence Diagram The Influence Diagram is where you define nodes and their relationships The bottom pane of the Model window is the Decision Tree which is covered later in this chapter Next you ll look at how DPL has established the links with Excel Double click on Gross Margin 147 Chap
218. e You will run an Option Value Chart using the Wildcat model from the previous section The process DPL uses for calculating option values will be discussed in the context of this example chart Click Home Sensitivity Option Value Chart The Option Value Chart setup dialog appears as in Figure 12 18 Option Value Chart Setup wildcat Decision _ P Evaluation method Save above settings as Default Alternatives Fast sequence evaluation Use Defaults defined in Node General don t show dialog again change Can be reset in Home Run Options Figure 12 18 Option Value Chart Setup Dialog 361 Chapter 12 Advanced Decision Analysis Results Syncopation Software This dialog is asking you to specify the default alternative states for each decision in your model for use in your Option Value Chart Note that the default states the dialog suggests are the same as the default states for each decision as specified on the General tab of the Node Definition dialog You can check the checkbox at the bottom left of the dialog if you wish not to be prompted for the defaults each time However when you use Option Value Charts you will usually want to examine and possibly change the default states You will change one of the default settings for purposes of generating the chart Use the combo box for the Test decision to change the default alternative to Core Sample Leave the default alternative for Drill
219. e a Z Paste g Sot amp Find amp z Por 08 98 E Cell Styles 7 Format 27 Filter Select Clipboard m Alignment a Number x Styles Cells Editing w M ir Wl M Mkt_ph2_probs A Initialization Links Probabilities and Values for Chance Nodes Marketing Phase 1 Low Nominal High Probabilities 0 3 0 4 0 Values 5000 8000 1500 1 2 3 4 5 6 7 8 9 Marketing Phase 2 Probabilities 0 3 0 4 0 Values 8000 10000 20000 Marketing Phase 2 Conditional Probabilities Marketing Phase 2 Marketing Phase 1 Low Nominal Low 0 5 0 3 Nominal 0 25 0 5 High 0 1 0 3 4 gt gt Assumptions amp Output Initialization CF 4 ia Ii Ready Average 0 333333333 Count 3 Sum 1 O A 100 Figure A 8 Excel Range with 3 Probabilities In DPL switch to the Data tab of the Node Definition dialog Select the first branch of the node gt Click the Links button a DPL displays a list of the cell names in the spreadsheet that you selected on the Links tab gt Select the appropriate range Click OK The range appears in the edit box as a reference in Sheet Cell notation See Figure A 9 448 Syncopation Software Appendix A Overview of Spreadsheet Linking Node Definition Marketing phase 2 General Conditioning Print Page Setup Copy Full Screen Wheel Reorder Multiple Expert Q q S Daa e Distribution type discrete y Links Probability
220. e you will see the get pay value for NPV below each branch of each Price node At the end of the tree the NPV value is for a single specific scenario You could verify the NPV for that specific scenario in the spreadsheet by setting the inputs in the spreadsheet to be the inputs for the path in the tree Since this model only has one get pay expression the rolled back expected value is the same as the value of the get pay expression at each endpoint which is why the same number appears above and below the branch You can use the Policy Tree to find out a number of things about your model For example the expected value of the model is 18 3 given that you develop the product in house and that sales turn out to be low 18 3 is the rolled back expected value above the node Cost per Unit on the Low branch of the Sales node which is on the In house branch of the Product Strategy decision 83 Chapter 3 Decision Analysis Results Syncopation Software gt Try manipulating the Policy Tree by expanding collapsing different parts of it You can do this by double clicking a node using the context menu or using the Expand Collapse split buttons within the Policy Display group on the ribbon You can also control the data elements that are displayed in the Policy Tree Select Policy Display Options the dialog box launcher with the Policy Tree active The Format Policy Tree dialog appears as shown in Figur
221. e 12 17 Policy Summary with Comparison Bars for Objective Function Filter The Policy Summary with comparison bars for the objective function being greater than or equal to 200 indicates that the Soaking outcome of Amount of Oil occurs in 100 of the filtered scenarios The comparison also indicates that the No outcome of Seismic Structure never occurs in scenarios where the objective function exceeds 200 You can also see that the Yes alternative of Drill occurs in 100 of the scenarios This information tells you that if Seismic Structure is in the No outcome there is zero chance that the objective function will exceed 200 In order for the objective function to exceed 200 you must drill and the amount of oil must be soaking 360 Syncopation Software Chapter 12 Advanced Decision Analysis Results Note that DPL gives each filtered Policy Summary a name that indicates which filter is active and how it is set You can rename or delete these comparisons using the Workspace Manager DPL will not overwrite them when you run the model and they are not cleared when you clear model memory 12 4 OPTION VALUE CHARTS Option Value Charts display the value of each option decision in your model In DPL the value of each option decision is defined as the difference in the objective function value when the model is run with that decision uncontrolled vs the value when the model is run with the decision controlled to a fixed alternativ
222. e 3 6 Format Policy Tree Display Branch length Branch offset 9 100 Show Node names State names Probabilities Get pay values Objective function expected values Attribute expected values Fat Policy Branch filter Based on probabilities Based on rolled back obj fr expected values Min 0 blin 0 Max Max 0 Figure 3 6 Format Policy Tree Dialog You can use the two spin buttons to increase or decrease the Branch length and Branch offset percentages to alter the shape and appearance of the tree gt Experiment with using the checkboxes to turn on and off the different data elements the tree displays 84 Syncopation Software Chapter 3 Decision Analysis Results 3 2 RISK PROFILE As shown in Figure 3 2 DPL creates an item in the Workspace Manager for each of the outputs it created To be exact DPL creates one or more items for each output For Risk Profiles DPL creates an item for the Risk Profile chart and an item for each Risk Profile dataset that is created In the decision analysis run that you requested previously DPL created one Risk Profile dataset for the optimal initial decision alternative If you had requested Initial Decision Alternatives DPL would have created two Risk Profile datasets one for In house and one for License In addition DPL would have created a Risk Profile chart showing the risk profiles for both alternatives In the Workspace Manager a Risk Profile chart is indicated b
223. e 6 8 189 Chapter 6 Asymmetric Trees Syncopation Software Develop se Yes Yes Nominal No No High Figure 6 8 Moving the FDA Approval Node Release the mouse button Figure 6 9 _ Low Low Nominal Nominal High High Figure 6 9 FDA Approval Node Attached to the Yes Alternative This model represents the same decision as the symmetric model but it has only about half as many paths 55 rather than 108 Runa decision analysis again The results approximately 75 million for the Yes alternative and 0 for the No alternative are the same Note that DPL assumed a value of zero for the No alternative since no get pay is specified on that branch Select the FDA Approval node in the Policy Tree gt Drop down the Policy Display Expand split and select Expand to Level In the Expand Policy Tree dialog type 4 Figure 6 10 190 Syncopation Software Chapter 6 Asymmetric Trees gt Click OK Expand Policy Tree Expand to level 4 2 6 oom full after expanding a Figure 6 10 Expand Policy Tree Dialog DPL expands the Policy Tree two more levels down enough to show you the expected values on the branches of Market Share Figure 6 11 Market Size Low 147 4 30 Market Share Market Size Yes 323 7 Nominal 213 4 50 40 Market Size High 942 0 FDA Approval 30 Develop Yes 74 8 i 74 8 Market Size Low 174
224. e Correlation Negative Correlation Figure 3 29 Value Correlations Chart The Value Correlations chart assumes that the optimal decision policy is followed It tells you which values in your model tend to vary in the same direction positive correlation or in the opposite direction negative correlation as the objective function which in this case is NPV Positive correlations are shown as purple bars while negative correlations are aqua A correlation bar is shown for each value in the model for which the correlation coefficient can be calculated Each value corresponds to either a decision node a chance node or a value node in the model If a chance node or decision node has no values associated with it DPL correlates the state numbers with the objective function E g for a decision node with two alternatives the first alternative has a state number of 1 and the second alternative has a state number of 2 As Figure 3 29 shows several of the values are highly correlated with the objective function The first bar NPV simply indicates that the objective function is perfectly correlated with itself Total Unit Sales Sales 1 Sales 2 and Price Change are all highly positively correlated with the objective function This makes sense because the higher the product sales are the higher the revenues and NPV are On the other hand Cost per Unit and Marketing Costs are negatively correlated with the objective function NPV License Fe
225. e High Lov Lew PABERI Wana m w Ye L s Mie Ma RERIG L w Lite Liw Stable High Lra Meenunal Pita Release 6 00 00 a WF Wea Lew Wie I Be EDM ID Leer Loe Low Stable High Love High Tanal Humber FYOPEDI IPERI J nn Ye Low Mit Ho UMID Loa Low Low Stable High Homin Loa Lipenied User Wubi Synicopaion Sor x z x TENS TSR0SR0080 Begin endecint aa z jak Wee Lida kia I Mi REM IO L s Lira Liw Shable High Hemin PMami s RHA iiia Begin eedonint A fo Yi Low File Ds Ma BCM IT Laver Lir Low Shable High Mierienal Hagh AES Ne Ered pipin dine i W TANI 1608 TTT Anyang Folin 2 254 Te Lew Bii I Ho BCH 10 Low Low Lin Stable High High Lra Poir diva gare IA eE Ve Liew Witi He DER IO Liw Levey Low Stable High Figh Mashina a aN Ves Low BCR Loree Lire Lim Stable High High High TES Tero orca i Flnceed time 365 Ps i Te Lew Mit Mie Bch i Lys Lon Low Unatable Lew Lor Lys i ee a Vex Low Mii Ho BEM Loe Low Low Unrateble Low Low Moeninal Li Pa ge ohne m 30 Tie Low Niet 1 Me REM ID Liw Lire Law Urdtikle Low L w High Persian zl jt Ves Lowe hile 1 No BDM 10 Lliw Lir Low Urdiable Low Meeneial Lire t iih Sieh 20n 14818 3 9033 1B bat n i 5 l Creda ts iD eae ISS Csatng nergy f gt ergy Medel E Gh imported ndpoint Maisha Wi aE Few Help press FE Coie H w ij 2 me Figure 11 26 Endpoint Database Imported from a CSV File As mentioned above DPL does not disp
226. e Influence Type Ea Probabilities Different probabilities for each conditioning event state Values Different values for each conditioning event state i Make this the default chance influence type Cancel Figure 7 10 Chance Event Influence Type dialog Click OK Now you are ready to enter data for Seismic Structure Double click on Seismic Structure The Node Definition dialog now has separate tabs for Values and Probabilities You don t need to enter values for this node gt the Probabilities tab Enter the probabilities as shown in Table Amount of Oil Seismic Structure Probability Noe fe o None 0 3 0 4 Closed 0 3 Soaking ope Closed oo Table 7 1 Probability Data for seismic structure 224 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models Your probability input tree should look like Figure 7 11 Node Definition Seismic Structure General Conditioning Print Page Setup Copy Full Screen Wheel Reorder qla Q Q Values 6 vif Links Seismic 6 Dry Structure Open a Closed a None Amount Seismic 3 of Oil Wet Structure Open 4 Closed 3 None Seismic 4 Soaking Structure Open Figure 7 11 Node Definition dialog Probabilities Tab for Seismic Structure gt Click OK An expert geologist gave you the probability distributions for different seismic structures given different o
227. e and Launch Costs are not correlated with the objective function at all because they are fixed deterministic values 111 Chapter 3 Decision Analysis Results Syncopation Software In decision analyses with complex value models Value Correlations charts are useful for understanding how the variability in the objective function is related to the variability in individual values such as chance nodes and downstream decisions 2 6 SAVING OUTPUTS All outputs described in Chapter 3 up to this point Risk Profile Charts and Datasets Policy Trees Policy Summary Value of Information Control and Value Correlations and Endpoint Databases see Chapter 11 for more information on Endpoint Databases will remain in the Workspace until the next run of the model When you have one or more of these outputs mentioned above associated with your model and you run a Decision Analysis you will be given a warning that any unsaved results from the previous run will be cleared or overwritten by the new run See Figure 3 30 DPL EJ This will clear any unsaved results for License ve In house Continue Don t ask again can be changed in Home Run Options Yes Mo Figure 3 30 Unsaved Results Warning An unsaved output or dataset will be cleared even if you do not select the output to be created in the subsequent run For example if you have an unsaved Risk Profile associated with your mod
228. e another attribute DPL names it Attribute2 by default See Figure 10 4 280 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics 4 Uo 4 Objective amp Utility Energy Model Objective Attributes NPV Risk Tolerance E Gather attribute expected values in the Policy Tree Objective function Maximize Minimize is w ly Units for output labels Constraint function Roll forward Roll back is w LA Figure 10 4 Objective amp Utility dialog with Attribute2 Added DPL automatically puts you in edit mode Type CF_2014 to rename Attribute2 Press Enter Click Add again DPL will add an attribute named CF_2015 which is what you want Click Add eight more times to create attributes named CF_2016 etc until you have created ten new attributes altogether the last of which will be named CF_2023 The Objective amp Utility dialog should look like Figure 10 5 281 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software Objective amp Utility Energy Model Objective Attributes Risk Tolerance E Gather attribute expected values in the Policy Tree Objective function Maximize Minimize Units for output labels Constraint function Roll forward Rall back Figure 10 5 Objective amp Utility dialog with Ten New Attributes Note DPL will automatically increment from the previous attribute name any name that ends with
229. e as ch Instance lt i Detach indicated by icon Use drop down Subtree split to select instance other than default Default changes to the last tree instance type added The dialog box launcher will open Default Tree rT Asymmetric the Tree Options Dialog where Asymmetric you can turn off on tree based Check the checkbox to make dependency checking change branches of selected node alternative grouping for new asymmetric Uncheck the checkbox nodes and turn off on advanced to make symmetric decision tree features Detach Detach selected node and create a new detached subtree starting with the selected node Subtree Create a Perform Subtree beginning with the selected node Table 1 4 Model Tree Ribbon Group 15 Chapter 1 Introduction Syncopation Software Control Combo Ctrl state Control All Control node to the branch selected Uncontro All in the combo or uncontrol if None selected Branch ig The dialog box launcher will open the Control tab of the Branch Control All drop down Definition dialog where you can Control all branches to Low first change branch control block or Nominal middle High last or set event states if advanced Default as defined in the node s features are on Other tabs definition state Use Uncontrol All to remove control from all branches in e Get Pay tab define a get pay the model expression for the branch e Risk Tolerance tab only ap
230. e expected value of all subsequent outcomes of chance events and the optimal selection of alternatives for all subsequent decision events Root Node The node on the decision tree that does not have any branches leading into it the first node on the far left of the decision tree Every tree contains a root node 4 3 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software S Sampling Approximating a full range of outcomes by evaluating a subset number of paths Scenario A path through the tree from the root to an endpoint that is a combination of specific decision alternatives and chance event outcomes This path a Scenario represents a single possible state of the world Scenario Risk Profile A Risk Profile in which the scenarios included in the Risk Profile are restricted to a particular subset of the optimal scenarios e g a particular combination of states of one or more events A Scenario Risk Profile can be specified from either the Policy Tree or via a dialog Schematic Diagram A type of decision tree diagram that does not show a full representation of the tree Most decision trees appear as schematic diagrams in the Model Window Sensitivity Analysis Allows you to investigate the impact of a node or group of nodes on the Expected Value and or Optimal Policy of a model There are many types of Sensitivity Analyses See Rainbow Diagram Two Way Rainbow Diagram Value Tornado Diagram Ba
231. e last node type you added gt Click anywhere in the Influence Diagram The Node Definition dialog appears Type Product Strategy for the Name of the node DPL will automatically split the node name into multiple lines Note there is an Alternatives list box on the General tab of the Node Definition dialog for a decision node You can add delete insert or rename decision alternatives By default DPL has created two decision alternatives and named them Yes and No These default alternative names can be changed within File Options General Also the first decision alternative Yes is designated as the default alternative You don t need to worry about this right now Default alternatives are relevant in some asymmetric models which are discussed later in this manual Select the Yes alternative type In house to rename it Note you may also click the Rename button or press F2 to edit an alternative name Rename the No alternative to License Your node definition dialog should now look like Figure 2 14 37 Chapter 2 Building Decision Analysis Models Syncopation Software DPL variable name for expressions Product_Strateay Altematives Figure 2 14 Node Definition Dialog for a Decision Node Click OK to close the Node Definition dialog Your model should now look something like Figure 2 15 38 Syncopation Software Chapter 2 Building Decision Ana
232. e objective function Double click the Risk Profile Chart item 4f labeled Expected Value in the Workspace Manager The Risk Profile Chart for the objective function is activated gt Note the maximum value on the x axis It is approximately 2500 Double click the Policy Tree item in the Workspace Manager to activate it again gt Click Policy Display Options The Format Policy Tree dialog appears as shown in Figure 12 3 In addition to allowing you to modify the appearance of the tree by changing the branch length and branch offset and by specifying which items to display the Format Policy Tree dialog allows you to filter branches 343 Chapter 12 Advanced Decision Analysis Results Syncopation Software Fomat Policy Tree Display Branch length honz Branch offset 100 Show iM Node names W State names Probabilities Fl Get pay values J Objective function expected values Attribute expected values Fat Policy Branch filter Based on probabilities Based on rolled back obj fn expected values Mir 0 Mir hax 1 D Canca Figure 12 3 Format Policy Tree Dialog 4 In the Branch filter section of the dialog check the Based on rolled back obj fn expected values checkbox 4 Leave the Min edit box in the rolled back objective function values section filled in with 0 4 In the Max edit box type in 2500 the value you noted from the Risk Profile C
233. e of the decision maker s attitude towards risk Common definition for risk tolerance r is the highest value for which you would accept a gamble in which you could win r or lose r 2 with equal probability In DPL the Risk Tolerance Coefficient describes the incorporation of risk when the built in exponential utility function is used Roll Forward The first phase of a DPL decision analysis run in which DPL calculates an outcome value and joint probability for each path in the decision tree As each path is traversed get or pay expressions are evaluated and combined to provide the path outcome value If the model uses multiple attributes each endpoint will have multiple outcome values Roll Back The third phase of a DPL decision analysis run in which DPL uses the outcome values and joint probabilities of each endpoint to calculate expected values or expected utilities at each node in the decision tree At each chance node in the Rollback procedure DPL determines the expected utility value by calculating the probability weighted average of outcome values At each decision node in the Rollback DPL determines the alternative providing the maximum or minimum value as appropriate The aggregate of these optimal decision alternatives comprises the optimal decision policy Rolled back Expected Value The value of the objective function at a node in the Policy Tree given the states of all preceding events on the path leading into the node and th
234. e programname The instruction may be coded in DPL programs and command procedures As with all DPL compiler directives this instruction must begin in the first column of the line In addition there can be only one space between include and programname When the compiler encounters the include directive during compilation or command execution it will suspend processing of the current input model or program and take its input from the program specified as the argument of the directive When the compiler completes the processing of the included program it will return to the suspended program and continue Included programs may be nested to any level however recursive include directives are not allowed If the compiler encounters an error in an included program DPL will place the cursor at the error location When compiling a program the DPL compiler will ignore duplicate variable definitions encountered in an included file and will use the first declaration of the variable found 422 Syncopation Software Chapter 15 DPL Programs and Code 15 6 3 Conditional Compilation Conditional compilation involves an if then else instruction for the compiler When the compiler encounters an if expression statement it evaluates the expression If the expression is true does not equal 0 it compiles the code immediately following until it encounters a else or endif statement If it finds a else statement it skips the code that follows
235. e project then FDA Approval does not matter the value is zero in either case If you expand down further into the don t develop half of the tree you ll see that all the values are zero This tree has 108 scenarios or paths to calculate this multiply the number of branches of the nodes 2 2 3 3 3 108 but half of them have an objective function value of zero These redundant paths increase runtime and make the tree harder to read 185 Chapter 6 Asymmetric Trees Syncopation Software gt Right click the Symmetric model in the Workspace Manager and choose Duplicate Right click the Symmetric copy model you just created Select Rename gt Rename the new model Asymmetric You are about to make some edits directly in the Decision Tree In the Decision Tree you may either have a node or branches selected In Figure 6 4 the decision node Develop is selected In Figure 6 5 the branches of Develop are selected Market FDA size Approval Pricing Low Low Low NPV Yes Yes Nominal Nominal Nominal No No NPV lt High High High Figure 6 4 Develop Node Selected in Decision Tree 186 Syncopation Software Chapter 6 Asymmetric Trees Market Pricin A size 2S Develop FDA Approval Low Low Low Yes NPV Yes Nominal Nominal Nominal mi SN 4 Ba High High High Figure 6 5 Develop Branches Selected in Decision Tree Double clicking on a node in the Decision Tree brings up
236. e sequence section you must use the integer keyword in that value s definition Don t confuse the integer keyword with the int function the int function truncates an expression value to the next lower integer the integer keyword rounds an expression value to the nearest integer 411 Chapter 15 DPL Programs and Code Syncopation Software You cannot define the values associated with a chance decision or controlled event to be integers The following code is incorrect incorrect code integer chance A 3 2 3 1 2 3 If you wish to define integer values conditioned by a chance decision or controlled event do the following correct code chance 4 43 2 3 integer value A given A 1 2 3 15 5 DEFINING EVENT SEQUENCES IN THE SEQUENCE SECTION The sequence section starts with the keyword sequence seguence If you are using multiple attributes defining an objective or defining a constraint function the word sequence will be followed by the appropriate specifications sequence attributes 2 objective 1 2 There is one important limitation on the organization of the sequence section keywords marking the states of a single decision or chance event must begin in the same column Apart from this you may define a sequence section with as many or as few lines as you like 15 5 1 Decisions You specify a decision in the sequence of events with the keyword decide or the keyword phr
237. e that DPL has changed the node data input tree so that it is conditioned on Product Strategy and has copied the data for Cost per Unit to each of these data entry slots Arrow down to select the License alternative gt Type 0 and press Enter Your node data input tree should look like Figure 2 23 This is a very simple model if Product Strategy is in the License alternative costs equal zero 4 Chapter 2 Building Decision Analysis Models Syncopation Software Node Definition Cost per Unit General Conditioning Print Page Setup Copy Full Screen wheel Reorder OSUEN ose C Links Cost Product In house per Unit Strategy 0 8 License 0 cant Heb Figure 2 23 Node Data Input Tree for Cost per Unit gt Click OK to close the Node Definition dialog You will now try a different procedure to create an influence arc between Product Strategy and Marketing Costs To do this you will select multiple nodes in the Influence Diagram by holding down Control and clicking Click Product Strategy to select it gt Hold down Ctrl and click Marketing Costs to also select it Click Model Influence Arc Add DPL creates an arc from Product Strategy to Marketing Costs If you have exactly two nodes selected DPL will create an influence arc from the first selected node to the second selected node when you click Model Influence Arc Add Now repeat the remaining procedure abo
238. e used to define a strategy i e an alternative may have no strategy icons beneath it indicating that the alternative is not considered in any strategy In this example every decision alternative is used at least once in some strategy 137 Chapter 4 Strategy Tables Syncopation Software m Node Definition Strategy General Commands Print Page Setup Copy Format Full Screen Data i Price Price Price Feature Feature Feature Marketing Set A Set B Set C Spend Change Change Change A B Cc Status Quo Add features Addfeatures Add features Increase Increase Increase Increase Pl Feature Led le HO P Price Cuts Keep same Keep same Keep same Keep Same Keep same Keep same Keep same Differentiate Product F Fo F Pip F O Minimal Minimal Minimal Decrease Decrease Decrease Decrease Figure 4 16 Strategy Window with Completed Strategy Definitions gt Click OK to close the Node Definition dialog gt Click anywhere in the Influence Diagram or press ESC to deselect the strategy node Your model should now look like Figure 4 17 Note that DPL has grouped the decisions added to the strategy table beneath the strategy node DPL has also updated the Default Tree in the Decision Tree pane It has removed the individual decisions and replaced them with the strategy table and the table s four alternatives strategies The Decision Tree now has 7
239. eadsheet Conversion See Conversion State One of a discrete number of possible settings for an event The states of a decision event are referred to as alternatives The states of chance event are referred to as outcomes Each branch of an event in the decision tree represents a state State Function A function that returns a numerical value representing the state of an event It can be used to process and test any event s state Statename Function Returns the name of a state of an event as a string High Low etc Strategy A single set of alternatives for all of the decisions contained in a Strategy Node Strategy Node A node which represents a set of decisions to be made at one point in time The different alternatives in a strategy node are called strategies The complete set of strategies in a strategy node is referred to as a Strategy table Each strategy is defined by an alternative for each decision included in the strategy node The graphical symbol for a strategy node is a yellow rectangle in the Influence Diagram with attached yellow rectangles for each decision node that is contained in the strategy node In the Decision Tree the graphical symbol is a yellow square During evaluation a strategy node is evaluated in the same manner as a decision node i e one alternative is chosen that maximizes or minimizes the objective function Strategy Table A collection of decision nodes and a set of defined strategies in a
240. ecision Alternatives Risk Profile for the energy case as shown in Figure 11 15 You can see from this Risk Profile that the two initial decision alternatives have very similar risk profiles and that neither alternative is substantially riskier than the other There is little reason for the decision maker not to choose the initial alternative with higher expected value which is High Initial Upstream Investment Low High 60 80 100 120 i140 160 180 200 220 240 260 280 300 320 Figure 11 15 Initial Decision Alternative Risk Profile from playing Endpoints 324 Syncopation Software Chapter 11 Using the Endpoint Database 11 2 6 Branch Controlling and Blocking You can very quickly generate what if analyses by controlling the outcome of an event to a particular state or blocking a specific branch of a decision and playing endpoints You should have a Policy Tree and a Policy Summary from the previous endpoint play with an expected value of 34 9 Right click on the Policy Tree and rename it to Original Policy Tree Right click on the Policy Summary and rename it to Original Policy Summary You have just saved the original results for comparison with your what if analysis You will evaluate how the model results change when two of the chance nodes are set to their worst case outcomes gt In the Decision Tree branch control the Actual Reserves chance node to the Low state You can do this directly on the
241. ed back expected value filtering That is if a branch is filtered out all subsequent branches further down the subtree will also be hidden from display regardless of whether their probabilities are in the range you specify 347 Chapter 12 Advanced Decision Analysis Results Syncopation Software gt Click Policy Display Options gt Uncheck Based on rolled back obj fn expected values gt Click OK to clear the filter 12 2 SCENARIO RISK PROFILES Scenario Risk Profiles can be used to view the risk profile for any particular scenario in your model A scenario is defined as a combination of fixed states of one or more events in your model Events that are not defined as part of the scenario are allowed to vary as in a Decision Analysis run generating a Scenario Risk Profile chart Scenario Risk Profiles use an Endpoint Database see Chapter 11 of this manual in conjunction with a Policy Tree so you must run Full tree enumeration and Record Endpoints in order to create them gt Make sure you have an Endpoint Database Policy Tree and Risk Profile from the example in the previous section If not create one of each now by re running the model Right click on the Risk Profile Dataset item called Expected Value gt Rename it Full Profile You have created a copy of the risk profile for the full model for comparison with a risk profile for a scenario Double click on the Policy
242. ed one of the outcomes Nominal as the default outcome You don t need to change this setting for the purposes of this tutorial Leave the settings as they are on the General tab and select the Data tab The Data tab for the Unit Sales discrete chance node is shown in Figure 2 32 Node Definition Unit Sales General Conditioning Print Page Setup Copy Full Screen Wheel Reorde Q Q Q Data Distribution type discrete v Links Probability 3 Value 20 lt wl Figure 2 32 Node Definition Dialog Data Tab for a Discrete Chance Node 60 Syncopation Software Chapter 2 Building Decision Analysis Models There are a number of elements on the Data tab for a discrete chance node You have already used the Conditioning button to condition Revenue per Unit The Conditioning button for a discrete chance node works in a similar way In addition to the Value edit box there is a Probability edit box and a Distribution type drop down list You can specify numbers or formulas in the Probability edit box in a similar manner to specifying data for values You can change the type of the distribution using the Distribution type drop down list The Distribution type drop down list will be discussed in Chapter 5 DPL assigns default probabilities to the three chance outcomes Like the outcome names these probabilities can be changed in File Options General DPL also assigns the low and
243. ed to use it later in this chapter J51 Chapter 12 Advanced Decision Analysis Results Syncopation Software 12 3 Povicy SUMMARY FEATURES As discussed in Chapter Section 3 3 the Policy Summary provides information about the probability weighted percentage that each decision alternative is selected in the set of optimal endpoints or scenarios For example in the model discussed in that section the Raise alternative of the Change Price decision is optimal in a probability weighted 30 of the scenarios You may wish to know in which scenarios the Raise alternative is optimal In that model the Risk Profile indicates that the objective function NPV in this case is negative in nearly 50 of the scenarios You may wish to know which scenarios lead to losing money DPL provides several features that allow you to compare the full set of optimal scenarios to those scenarios filtered on specific criteria These comparison features help you find out the type of information alluded to in the above examples This section covers the Comparison filtering features of the Policy Summary output Select File Open Navigate to the Examples folder underneath where you installed DPL If the default location was used for installing the path is C Program Files x86 Syncopation DPL8 Examples Select Wildcat da DPL opens the Workspace for the Wildcat model as shown in Figure 12 11 352 Syncopation Software Chapter 12 Advance
244. eeeeeeeeeeeeees 56 SQUUP siistnotenstnenecimicpnvanuciemtennam 53 Workspace ManaGef ssssscsseeeeeeees 6 NON WINCOW ItEMS cceceeeeeeees 85 WWW SYNGCOPATION COM
245. eessesereeeseees 194 Default Tree seceessrsereesennes 39 103 Cefault CASC ccccsssresceessreeeeeenes 405 Definition section s es 393 399 discrete chance node csceeeeeeeee 59 default probabilities 00008 61 default values cccccecceeeeeeeeeees 61 Document Navigator sses 9 dont keyword sssssssssssrsrenrsrerenn 417 downstream CeCISION 0sceeeeeeees 91 downstream decisions elei e a 203 and real OptiOns ssceeeeeeees 203 DPL Code sssaaa See DPL Program DPL Program ssssssssessrsrsrsrerererenns 391 DPL variable Names 43 DPLMC DEL sisinta 401 edit UNG O sisisi 317 Endpoint Database scceeeee 305 EXD OFAC drsna riana 330 FICE niona 310 IMPONG csctsevinewtinsevbineneennannaes 334 MEIGiNG ccscceceeceeseeeeeeeseeneseeaes 338 Piaya 315 RECONNECTING cccsceeeeeeeeeeeeees 327 Recording cccscceecesceeeeeeeseeees 306 SAV INCE E 322 SOMUNAG PO E E 314 state names numbers nsss 331 error DAMS ccesessesceueereeeeneureeseneanes 291 event tOrNACOES ccecceeeeeeeeeeees 384 CEterMiINIStiC ccecceeeeeeeeeeeees 384 probabilistic siaceravierveresencssseneerss 384 excel CIAUSE ccececeeeeeeeeeeseeeeseees 400 EXIT ODEON ssaranamacterrsinesernecrnnn 203 expected value of perfect information COMUOL seisiascavsissnttescesserscanssens 107 Fat Wr OUCY wateceinetnotaseiiivensesaeasnaness 245 Index Gamble keyword sssseeseeseeeees 413 Get keyw
246. efinition of a user defined utility function equivalent to the built in exponential utility function 372 Syncopation Software Chapter 13 Risk Tolerance Objective amp Utility License vs In house prob Objective Builtin exponential utility function Risk tolerance coefficient ly Fi Risk Tolerance You can use either the builtin exponential utility function or a custom function but not both A custom function is only necessary when the exponential utility function is inappropriate F you use a custom function you must specify functions to convert both to and from utiles e the utility function and its inverse Custom utility function To utiles xp Risk_Tolerance From utiles Risk_ Tolerance log Figure 13 8 User defined Utility Function Note Most applications that require modeling risk aversion can be done so by specifying a single risk tolerance or utility function for the entire Decision Tree However there is a theoretical argument for changing the degree of risk aversion throughout the tree as the company s situation changes DPL provides this capability although it is hidden by default since it is seldom used To use this feature check the checkbox next to Risk tolerance and utility functions by branch in Model Tree Options under the Advanced Setting section The Branch Definition dialog will then have a Risk Tolerance tab on which you can set either a risk tolerance or a custom utility func
247. egend by using the commands available in the Chart Format Legend group If you would like it to be placed as it is in Figure 5 24 choose Right from the drop down list and check the checkboxes next to Centered and Inside 167 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software It s clear that the three risk profiles are similar in a general sense but that the 100 Samples curve deviates from the other two This deviation may be easier to see if the lines were thinner gt To decrease the weight of the line of the curves select a curve it should turn magenta and go to Chart Series Lines Markers Within the Weight edit box delete 2 and enter 1 Do the same for the other two lines See Figure 5 24 100 95 90 85 80 75 j 70 f 65 f 60 a O 55 O o P 50 100 Samples 2 1000 Samples 45 10000 Samples E 40 O 35 30 25 20 i 15 j 10 5 0 OO O M N M LDL 800 600 400 200 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Figure 5 24 Formatted Sample Size Comparison with Legend If you like you can compare the statistics and percentiles of the risk profiles by clicking Data Distributions Statistics again Use the drop down list to select from the three distributions You may enter other percentiles into the Percentile list box and then click Update Percentile Values 168 Syncopation Software Chapter 5 Buil
248. egyTableSubtreel gamble on Market 2 Political Stability Future then perform a to Pipeline Strategy Mkt I na 20 BCM and 396 Syncopation Software Chapter 15 DPL Programs and Code set Pipeline Capacity BCM 20 and Set Include Market 2 Yes Aand perform StrategyTableSubtreel tO Pipeline Strabeqy Mkt 1 jands2 30 BCM and set Pipeline Capacity BCM 30 and set dnclude Market 2 Yes aad perform StrategyTableSubtreel to Proceed No and get Upfront NPV Spend a few minutes examining the program you have created Below are some of the characteristics of the code that you should notice The first line of the program tells you that the model is linked to a spreadsheet and gives the spreadsheet s filename and path DPL gives the spreadsheet the label Excel_1 and the value chance and decision nodes that are linked to this spreadsheet are preceded by the excel keyword and the name of the linked cell for example excel Excel 1 Assumptions Market 1 Size value Market 1 Size 110 The majority of the code is the definition section representing the Influence Diagram as described earlier See Section 15 4 for more information on the definition section DPL automatically inserts comments the text following each indicating the settings of each event state whose probability and or value is being defined for example Actual Price Low Nominal High 3 4 3 ioy Actual Price Low 254 Actual Price
249. el and run a Decision Analysis in which you do not request a Risk Profile output the Risk Profile from the previous run is still deleted To save outputs from a particular run so that they will not be overwritten in subsequent runs you must re name the output in the workspace manager Note that there is no limit to the number of saved sets of results you can have in DPL workspace Furthermore any formatting done to those outputs will be saved by re naming 112 Syncopation Software Chapter 3 Decision Analysis Results Outputs generated from a sensitivity analyses tornado diagrams rainbow diagrams Time Series Percentiles and Option Value charts are not overwritten in subsequent Decision Analysis runs or analyses They will remain in the workspace until you explicitly delete them When you save a DPL workspace File Save you will save all outputs of your of workspace including any formatting done 3 7 RAINBOW DIAGRAM ON A VALUE The rest of this chapter addresses Rainbow Diagrams which are another form of sensitivity analysis A Rainbow Diagram provides more detailed information on how the objective function of the model and the optimal policy change as a single variable is varied across a range of settings You will now create a Rainbow Diagram for a value node in the model gt Click Home Sensitivity Rainbow Diagram One Way Rainbow diagram in is the current default indicated by the icon The Run Rainbow Diagram di
250. elect Current value 0 06 From D3 Humber of steps 4 Step size 01 Evaluation method Monte Carlo simulation Initial samples 5000 Restart samples 100 Cancel Help Figure 5 34 Run Rainbow Diagram dialog gt Click OK to run the Rainbow Diagram 1 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software DPL will run a full simulation for each of the four steps 0 03 0 04 0 05 0 06 a total of 40 000 recalculations of the spreadsheet This may take a few minutes The Rainbow Diagram Figure 5 35 shows the model s expected value for each of the steps The color change between 0 04 and 0 05 indicates that the best highest expected value decision alternative changed between these two points This tells you that if you knew Loss of Synergy would be only 4 it would be better to spin off Enlightify If you wanted to find the threshold more precisely you could run another rainbow diagram with a range of 0 04 to 0 05 and a smaller step size Vv 2 o gt U wv bo U D Q x lt i Loss of Synergy Figure 5 35 Rainbow Diagram on Loss of Synergy This Rainbow Diagram only has two colors and the model only has one decision with two alternatives so it s not hard to figure out the alternative represented by each color In a more complex diagram you can use DPL s policy tip feature to see the optimal alternative for each step If you hold the mouse cursor
251. ependent probability of the optimal branch will always be equal to one or 100 and will be zero 0 for the rest So in the Policy Summary the optimal decision alternative for each initial decision is the one with 100 probability As you know In house is the optimal alternative and Figure 3 26 displays In house with a 100 policy dependent probability For downstream decisions the Policy Summary displays the probability weighted fraction of scenarios for which each downstream decision alternative is optimal In your updated model Price Change is a downstream decision because it occurs after the Sales 1 discrete chance node As indicated in Figure 3 26 Raise has a policy dependent probability of 30 while Keep same has a policy dependent probability of 70 By looking at the Policy Tree again you can determine in which scenarios the Raise vs Keep same alternative is optimal The Policy Summary also displays the policy dependent probabilities of the outcomes for discrete chance nodes For discrete chance nodes that are not conditioned by others and that do not depend on the optimal policy the policy dependent probabilities are equal to the input probabilities This is true of Sales 1 Cost per Unit Marketing Costs and Price However Sales 2 is conditioned by another node namely Sales 1 You can see that the policy dependent probabilities for Sales 2 are not the same as the input probabilities 106 Syncopation Software Ch
252. er results expected value of 34 9 because your model is still linked to the original Excel file gt Before proceeding to the next tutorial switch to the Energy Model and Home Run Clear Mem to clear model memory Press Ok to the warning The imported Endpoint Database is no longer available 11 5 2 Importing Endpoints from CSV Files In this section you will see how to import the Endpoint Database from a CSV file that was exported from DPL The steps are simple gt With the Energy Model activated and model memory cleared select Home Workspace Add to WS External Endpoints Browse to the folder where you installed DPL using the Files of type drop down list to select Comma Separated Value Files CSV Select endpoints_import csv from the list Click Open In the Import Endpoints dialog check Display report so that you will be able to see the imported endpoints this time 4Y OGY 4 Click OK DPL displays the endpoint report and calls it Imported Endpoint Database as shown in Figure 11 26 336 Syncopation Software Chapter 11 Using the Endpoint Database 8 Guo Grd DPL Emengy Case imparted Endpemnt Database e Home Belicocie i view Din anid Help 4 Te fy P 1 Ba E Se a a3 Sort Fiter Clear J Amaan Columns Tj Wiorkppape bianigrr p o I i a 4 5 7 a Lo 11 1 li Ey A sjak Lnengy Case 551da O Endpoint Proceed bhim U Pineline_ Include Pipeli
253. erance Place the node near Profit Name it Risk Tolerance Click the Data tab Type 50 for the value Click OK 368 Syncopation Software Chapter 13 Risk Tolerance gt Within the Model Objective amp Utility group there is a Risk tolerance combo box Select Risk_Tolerance from the drop down list See Figure 13 4 Max Objective Function Risk tolerance Risk_Toleranc W Objective amp Utility Figure 13 4 Risk Tolerance Combo Box You might ask where the 50 comes from In rough terms it s the amount you or your company are willing to put at risk You can assess your risk tolerance coefficient directly using the following rule Imagine that a fair coin will be tossed If you call it correctly you win R if not you lose half of R What is the largest R for which you would be willing to play this game The largest R for which you would be willing to play is the risk tolerance coefficient Risk tolerance coefficients are difficult to assess and are never very precise You should always test your risk tolerance coefficient using sensitivity analysis as you will below For this reason it is good to define a value node for your risk tolerance rather than specifying the number directly If you did the latter you would not be able to run a sensitivity analysis on it You ll now run the model and observe how the results have changed Select Home Run Decision Analysis gt Click OK for the warning
254. erent alternative on each computer You could then quickly conduct a complete analysis after merging the separate Endpoint Databases In this example you will merge two databases that were created using the energy model Each database contains half 16 056 of the endpoints in the overall database so each CSV file contains a comment section and required formats at the top followed by the column heading row and then 16 056 rows of data If you like examine the files endpoint merge 1 csv and endpoint merge 2 csv which are in the Examples folder beneath where you installed DPL Close the CSV files Switch to DPL Make sure that the energy model is open and activated and has no Endpoint Database in it Press F9 to clear model memory if needed 4 Gedy Select Home Workspace Add to WS External Endpoints Browse for the file endpoint merge 1 csv you will have to use the Files of Type drop down to look at files with the extension csv as you did earlier Open the file In the Import Endpoints dialog keep Display report unchecked Select Endpoint Database endpoint imerge 1 and click OK DPL loads this database gt Select Home Workspace Add to WS External Endpoints again This time locate and open the file endpoint merge 2 csv 338 Syncopation Software Chapter 11 Using the Endpoint Database gt In the Import Endpoints dialog check Merge with existing endpoints as shown in Figure 11 27 Clic
255. ery When a utility function is specified the certain equivalent is provided in addition to the expected value Chance Node A node in the Influence Diagram or decision tree that represents an event whose outcomes are uncertain Chance nodes can be either discrete or continuous Chi Distribution A named distribution supported by DPL For a detailed description see On Line Help Chi Square Distribution A named distribution supported by DPL For a detailed description see On Line Help Code A model or part of a model written in the DPL programming language Can be viewed in a Program Window Command Window A testing and debugging environment in which you can display or temporarily change any value or calculation in a model Compilation Before running an analysis on a model DPL compiles the model During compilation DPL checks for syntax and other errors and prepares the model for evaluation by the DPL engine Conditional Decision Policy A policy containing decisions which will be made after initial uncertainties are resolved The optimal alternative may depend on the state of the uncertainties and may therefore be different in different parts of the decision tree Conditioned When a value or event has separate numbers or formulas for either its values or probabilities or both depending upon the states of one or more other events For example if Costs 0 if Develop Drug is No and Costs Development Costs Launch Costs if De
256. es for points in time in between the discrete time periods are not known However you may find this depiction of the percentiles more useful than the error bar depiction for communication purposes Also note that within the Chart Format tab you can change the Titles Subtitles Axis labels Legend marker line color and background display of the Time Series Percentiles chart To format the font within the chart select the font and use the commands in the View Font group Try this if you like See Figure 10 14 for an example of what the Time Series Percentile dialog looks like when you specify that two of the percentiles as the minimum and maximum of a range bar 292 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics Time Senes Percentiles Energy Model Time periods attributes Percentiles From CF_2014 Type To CF 2023 oak p ange Min Initial time period 53014 Faint for x axis label Range Max Paint Paint Paint Aun settings Evaluation method Full tree enumeration Number of intervals 500 Cancel Figure 10 14 Time Series Percentiles Dialog for Range Bar The settings in Figure 10 14 produce the Time Series Percentile Chart below in Figure 10 15 You can see that the 95 percentile of the cash flows is greater than the 90 while in most time periods the 5 percentile is the same as the 10 As specified the 10 and 90 percentiles are displayed as the minimum an
257. es should be used for each attribute For more information on specifying Get Pay expressions for multiple attributes see Chapter 8 as well as the last section of this chapter gt Double click the branches of the CapEx node in the Decision Tree The Get Pay tab of the Branch Definition dialog appears as shown in Figure 10 7 283 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software Branch Definition Get Pay CapEx Get or pay expressions Control OA itl a co NPV Total_NPV Display aaa aac aaa aaa i CF 2015 CF_2016 amp CF_2017 CF_2018 CF_2019 CF_2020 CF 2021 CF_202 CF_2023 Go to Attributes Get Pay LI ivi f Figure 10 7 Get Pay tab of Branch Definition Dialog Now that there are multiple attributes defined in the model DPL provides room in the Get Pay expression for separate expressions for each attribute When you have multiple attributes in a model you must specify a value for each of them in every Get Pay expression or use zero as a placeholder The Get Pay expression as it is currently defined only has a value Total_NPV for the first attribute called NPV If you had multiple value nodes defined in your model for each attribute you would need to use them in each separate expression in each of the ten get pay rows in this dialog However your model is set up so that these ten attributes are determined by a single 1 x 10 value node gt Click in the expres
258. es you the ability to model up to 1024 attributes Some or all of these attributes can be weighted and combined in an objective function DPL selects optimal decision alternatives based on objective function values maximizing or minimizing as appropriate The objective function should not be confused with a utility function for risk tolerance see Chapter 13 The objective function combines several financial and or other attributes into a single value whereas a utility function typically adjusts a single financial value to account for risk aversion Zoo Chapter 8 Objective Functions and Multiple Attributes Syncopation Software S 1 INCORPORATING MULTIPLE ATTRIBUTES You ll begin with a model that has only a single economic attribute You ll add a second attribute to this model to account for environmental considerations gt Open the sample file MUA da in the Examples folder underneath where DPL is installed C Program Files x86 Syncopation DPL8 Examples if the default location was used during installation gt Double click the splitter bar between the Influence Diagram and the Decision Tree to maximize the Influence Diagram See Figure 8 1 DPL MUAde Wildcat oE Figure 8 1 Influence Diagram for the Single Attribute Wildcat Model gt Make sure Policy Tree is checked gt Click Home Run Decision Analysis or press F10 Based on the single attribute of profit the optimal decis
259. eters chance A High Medium Low beta 1 2 For more information on named distributions refer to the online Help Chance events may also be continuous A continuous chance node is defined in two parts one for the chance event and one for the associated value chance B 1 value B DPLMC normal 10 3 The first line above defines a chance event with a single unnamed state this is how DPL knows the event is continuous discrete events have two or more states The second line tells DPL that the chance event will be drawn from a normal distribution and that the mean and standard deviation will be 10 and 3 respectively DPLMC refers to DPL s Monte Carlo library DPLMC DLL which is part of the DPL installation For a list of available continuous distributions see the online Help If the probabilities or values associated with a chance event depend on the states of other events you include this information in the chance event s definition You must define the conditioning events before you define a chance event that depends on them You must provide probability distributions and value expressions for each combination of conditioning event states 401 Chapter 15 DPL Programs and Code Syncopation Software For example the following statement defines a chance event that depends on a decision called Research_Funding and a chance event called Research_Time both defined earlier Both Research_Funding and Research_Time are condi
260. eting Costs variables are not included in the table initially because they are conditioned by the decision and thus have more than one value We will add these variables to the table manually gt Within the Value group select Cost per Unit from the drop down list and click Add A line for Cost_per_Unit Product_Strategy In_house will be added to the table Again within the Value group select Marketing Costs from the drop down list and click Add Click the blank cell for the Low value of Launch Costs and enter 20 Click the blank cell in the High column of Launch Costs and enter 25 LUUL Q Continue to fill in the Value Tornado Setup dialog with the rest of the High and Low values shown in Table 2 2 Note that you can use the arrow keys to move around 53 Chapter 2 Building Decision Analysis Models Syncopation Software Launch Costs 20 25 Table 2 2 Variables and Ranges for a Value Tornado When you are done the Value Tornado Setup dialog should look like Figure 2 28 Value Tornado Setup License vs In house Bars Value Low Current High Launch_Costs 20 22 25 License _Fee 0 7 0 75 0 8 Price 1 9 2 2 3 Unit_Sales 20 4 60 Cost_per_Unit Product_Strategy In_house 0 8 0 8 1 Marketing Cost Product Strategy In house o5 Remove Yehi Current value Price j 2 Add Replace Conditioning _ Initial Decision Alternatives Figure 2 28 Value Tornado Setup Dialog 54 Syncopation Software Chapter 2 Building Deci
261. etween the node and Excel There are two categories of spreadsheet links calculation links and initialization links Calculation links are used to send or receive data from Excel during the course of a run Initialization links are used to set the values or probabilities for a node at the beginning of each run Initialization links are discussed in Appendix A 3 The radio buttons near the top of the Calculation links section of the dialog indicate whether the node is linked and if so to what Excel or a DPL Program The workbook as well as the sheet and cell that the node is linked to are displayed in edit boxes The sheet and cell syntax is the same as Excel s syntax i e Sheet1 Cell_ name The table below summarizes the functions of the buttons in the Calculation links section Button Action Browse Finds a spreadsheet or allows you to change to which spreadsheet this node is linked Brings up the Cell Names dialog which can be used to change the sheet cell to which this node is linked Takes you to the cell in Excel to which this node is linked Pastes a link to a cell from Excel Table 2 1 Calculation Links Buttons Click the Go to Cell button DPL activates Excel and selects the sheet and cell to which the node is linked As you can see in Figure 2 8 the cell named Cost_per_Unit contains the value 0 80 which is what DPL copied for the node s data 30 Syncopation Software Chapter 2 Building Decision A
262. fault for now Attribute2 through Attribute6 You are setting up Get Pay expressions on the Proceed branch of the tree Initially the Get Pay tab of the Branch Definition dialog for this branch would look something like Figure 10 21 300 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics Branch Definition Get Pay Proceed Get or pay expressions Control Attribute Get pay expression NPV Attributed Attribute3 Attribute Attribute Attribute Display Go to Attributes Figure 10 21 Get Pay tab of Branch Definition Dialog for a Six Attribute Model You need to define values for each of the five attributes other than NPV which is already defined as Total_NPV There are several ways in which you might use a multidimensional value node to fill in get pay expressions Assume first that you have defined a 1 x 5 row value node called Invest_1x5 as shown in Figure 10 22 301 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software Node Definition Investment_1x5 G ji al i E i mm at Columns 5 E Seres interval entry Z Create M x 1 column vector lv f Column Links 0 Figure 10 22 A 1 x 5 Value Node Definition In order to use this value node to fill in the Get Pay expressions on the Proceed branch you would take the following steps gt Double click the branch to open the Get Pay tab of
263. ferent sets of input values drivers to Excel and have Excel send back one or more output values metrics Nodes in DPL that send data to Excel are called export nodes Nodes that receive data back from Excel are called import nodes Export nodes are typically linked to cells in Excel that contain constant numbers not formulas Import nodes in DPL are typically linked to cells in Excel that contain formulas In our model NPV is an import node and the other nodes are export nodes gt Double click on the Cost per Unit value node The Data tab of the Node Definition dialog is displayed See Figure 2 5 2 Chapter 2 Building Decision Analysis Models Syncopation Software Node Definition Cost per Unit General Conditioning Print Page Setup Copy Full Screen whee Reorder SEIR vis A Links Cost per Unit o 0 8 Figure 2 5 Node Definition Dialog with Data Tab Selected The Data tab displays a data input tree for the value node and allows you to enter the data for this node Because this node was created as a linked value DPL has filled in the value of 0 8 which comes from the cell named Cost_per_Unit in the spreadsheet There are several buttons on the Data tab which allow you to change items such as the node s conditioning to view things differently by changing the zoom or going to full screen or to help you input the data for this node by selecting variables We ll discuss these later gt
264. fine Switch back to DPL Click the Links tab of the Node Definition dialog Under Calculation Links select Microsoft Excel Click the Cell Names button In the Range Names dialog check the checkbox for Show ranges with formulas The range you defined in Excel currently has a formula in it Select Mkt_1_All_ Years from the list You can see that it has a Count or dimension of 25 cells Click Select Click Yes for the warning DPL is telling you that you have just defined an export node that will overwrite a formula in Excel The Links tab of the Node Definition dialog should now look like Figure 9 14 269 Chapter 9 Multidimensional Value Node Syncopation Software General Data Node Definition Mkt 1 All Years Calculation links These links are recalculated throughout the run They may be either inputs used in the linked spreadsheet s calculations e g unit sales or calculated results from the spreadsheet e g NPV None local DPL Program Microsoft Excel Workbook Multidimensional Value xs Browse Sheet Cell Assumptions Mkt_1_All_Years Go to Cel This is an export node Its value will be sent to Excel each time it changes during the run Paste Link Initialization links These links are for connecting to a spreadsheet or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at t
265. g Values gt Click OK to run the Two Way Rainbow Diagram DPL produces the Two Way Rainbow shown in Figure 14 7 DPL creates a Two Way Rainbow Diagram by running the model for each combination of the two values selected for the diagram To create the Two Way Rainbow Diagram in Figure 14 7 DPL ran the model 99 times 380 Syncopation Software Chapter 14 Sensitivity Analyses 0 9 0 875 0 825 License Fee gt z 0 725 0 8 0 825 0 85 0 875 0 9 0 925 0 95 0 975 1 1 025 1 05 Per Unit Costs Per Unit Costs High Figure 14 7 Two Way Rainbow Diagram A Two Way Rainbow Diagram displays policies using colors and marker types for the values on the x and y axes The policy at all points with the same color and marker type is the same A color marker change indicates a change in policy The Two Way Rainbow Diagram in Figure 14 7 indicates that the optimal policy depends on both the value of Per_Unit_Cost High and License_Fee This can be seen because the change in policy as License_Fee increases occurs at lower and lower values as Per_Unit_Costs High increases For example when Per_Unit_Costs High is 0 8 0 825 or 0 85 a change in policy occurs as License_Fee increases from 0 85 to 0 875 but when Per_Unit_Costs High is 0 875 a change in policy occurs as License_Fee increases from 0 825 to 0 85 To see which policy is optimal at each point and what the expected value of the model is at each point in the Two
266. ge in the tree For completeness click Model Influence Arc From Formulas You may want to move some nodes around again and or bend some influence arcs to make the Influence Diagram neater Your model should look something like Figure 3 25 You will now run your updated model Marketing Price rice Product Cost Costs ot Strategy Change per Unit coe house Raise Coon Nominal Nominal Nominal Nominal BGs E Keep Same Or iy A High High High ie Figure 3 25 Completed Model with Downstream Decision gt In the Home Run group make sure the following are checked Risk Profile Policy Tree Policy Summary and VOIC gt Click Home Run Decision Analysis or press F10 DPL produces the requested outputs Review the Policy Tree before looking at the Policy Summary 104 Syncopation Software Chapter 3 Decision Analysis Results In the revised model we have a downstream decision With downstream decisions there is not necessarily a single alternative that is always optimal Note that in the Policy Tree a different alternative is optimal for the Raise Price decision depending upon the outcome of Sales 1 It is easy to review the relatively small Policy Tree in this example The Policy Summary provides a useful view of the information in the Policy Tree particularly for models with larger and or more complex Policy Trees Note that the expected value of the model has increased to 111
267. gether Laplace Distribution A named distribution supported by DPL For a detailed description see On Line Help Links Connections to spreadsheets databases or DPL programs that allow DPL to perform calculations or obtain data Local Variables Nodes that are not linked to a cell or cells in a spreadsheet to a record in a database nor to a DPL program Logistic Distribution A named distribution supported by DPL For a detailed description see On Line Help Lognormal Distribution A named distribution supported by DPL For a detailed description see On Line Help 468 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms Lottery An event whose outcome depends on chance M Marginal Probability The probability that a given outcome of a chance event will occur For example if there is an even chance of sun or rain the Marginal Probability of rain is 0 5 Maximize Instructs DPL to choose the decision policy that returns the maximum expected value for the Objective Function Maxwell Distribution A named distribution supported by DPL For a detailed description see On Line Help Minimize Instructs DPL to choose the decision policy that returns the minimum expected value for the Objective Function Model Window Graphical interface for designing decision models and running analyses Contains both the Influence Diagram pane and the Decision Tree pane Modified Monte Carlo Simulation method of ev
268. gh High High Reece 8 D3 lt f Dore T 7 Serial Number SYDEON 1 Breakthrough PW Market Licensed Us io nc Hana 1t lt gt lt gt For Help preys FI icom 10 Full Prev l Figure 12 1 Policy Tree da This is a drug development example gt Make sure Risk Profile Policy Tree and Endpoints are checked in the Home Run group Make sure that Init Dec Alts and all other outputs are unchecked gt Run a Decision Analysis DPL launches Excel if it is not already running and loads Policy Tree xls runs the model and produces the requested outputs The Policy Tree as shown in Figure 12 2 will be active 342 Syncopation Software Chapter 12 Advanced Decision Analysis Results Market Share Launch Yes 400 4 Failure 400 4 122 9 30 No 174 0 Market Share Key competitor outcome Launch _Yes 221 0 FDA Approval _Yes 177 9 Moderate succ 221 0 122 9 Develop _Yes 2 0 50 40 No 174 0 2 0 174 0 Market Share Breakthrough 102 0 122 9 30 No 174 0 No 174 0 50 No 0 0 gt Figure 12 2 Policy Tree for Drug Development Model You can use the Format Policy Tree dialog to prune some of its branches Suppose you are interested in seeing the scenarios in which the objective function the net present value of the cash flows in this example is positive To do this you first need to have an estimate of the maximum value for th
269. graphics e Modernized icons e Excel side interface e New Tornado Diagram type called Probabilistic Base Case which helps you to better understand the risk drivers in your model e The ability to fully format and customize your outputs within DPL no need to export and fiddle to get your outputs presentation ready e Node collections that allow you to group nodes into collections and collapse them for condensed display in the Influence Diagram e Find replace in the Influence Diagram and Decision Tree e Ability to use relative paths for spreadsheet links so that you can more easily pass a DPL file and Excel file combination to a colleague e Ability to embed your DPL Workspace in an Excel file so your model is in a single file The above features and many other new features are demonstrated and explained throughout the chapters of this manual 22 Syncopation Software Chapter 2 Building Decision Analysis Models 2 BUILDING DECISION ANALYSIS MODELS This tutorial focuses on building a spreadsheet linked DPL model for a simple decision analysis application It will familiarize you with key DPL features by providing step by step guidance on how to create a DPL model from a simple Excel spreadsheet modify the model and generate results If you are new to DPL this tutorial is a good place to start A DPL model is a combination of an Influence Diagram and Decision Tree The main purpose of an Influence Diagram is to show the relations
270. gt Click Yes gt Click OK for the warning Right click on the Endpoint Database and Rename it to Saved Endpoints DPL makes a copy of the Endpoint Database under the new name See Figure 11 18 Jid oa DPL gt Energy Cande ined Endpin 5 al iH ja al FA SE LH pt ra Coun 7 Workpace Managert i 1 4 J 4 i 1 4 yo Weiipeia Oi Eniponi Freed eal UL Ppeti leche Popeiee_ Capi Aue Pa Atul Ro Marina Porsraako Mie Operate Apereda Prebabe Depectet_Fu gt ee Encigy Mesiel aa a o Love Wai He BCHM I Loe Low Le Stable Liew eww Lew Doh 16 F Undpetord Draka 7 1 Vei Lew ECT He BEMI L r Lew Lew Sieber Lew Lew Harr DETA 160 i Std Eediprerts rae F 3 5 Y Ligne kiii T Ho BCM_t Low Low Low webs Low od tgh 16 0 i Low hia He BCA Lins Lin Lee Stable Lew fori Liw 169 4 Low Miir Ha BCH I L r L n Low nabit Law memi Harm 16 b l F Wia 1 Pic DOME Low Low Low b Low ee The 3 lpw kiia He BCH IJ Lim Lens Loew Stable Lew High A 1409 i r Lave hia He BOMLES Lew Lew Low Seabie Low Hgh Harm TED 7 E Ler Hi T e He DENT Lya kga ksa mat baw agh Pgh The i Live klia He BEHI L s Ls Loe ak Herren Baa Ea 18E ii Lira Wiig He BCM BS Low Lew Low Spb Hiat Low Harumi TED Iz li Ye Low Wiii Ho BCH Lya Low Low Stable Piorraud Lo Fgh oS J D EE i Low hii He BEMIS Lo bow Lyw b minai lominai ETE ta 1 Lira Wa He BOM I L w L n Le Snabhe Hormia Homma Harmina 16 0 it 13 Ye L Miil Me BLA
271. has been assigned a range name When using an Excel model with DPL you must name the cells you intend to link to DPL If you select a named cell in Excel the name will appear in the name combo box to the left of the formula bar In Figure 5 3 you can see that cell C17 has a current value of 895 and has been given the name Total_Value Please consult your Excel documentation if you are not familiar with named ranges as they are essential for using DPL with Excel Z ia fy als Compatibdlty Mode Microsoft Exc tome insert Page layout Formulas Data Revew ew oA B A Arial T Cuvengy Ei Conationsi formating g insert Br 3 E SJ J amp romsa ndes Delete M DOE u i 2a pg 23 Cen Stytes gt 13 Format Cip oso Font 1 Total_Value nd s ae Abgnment Se mmdet Stytet Com fa NPV SS Enlightify Consulting LLC oo nr eerk WwW fa gt Let ts M a O D ses _ Figure 5 3 Cell C17 is Named Total_Value The calculations sheet DCF is shown in Figure 5 4 In this sheet the model calculates the total value of the Enlightify Consulting business unit by the discounted cash flow method These calculations are connected to the input cells on the Assumptions sheet For example if the Trendline Revenue Growth Rate cell C22 is changed from 8 to 5 the total value of Enlightify falls from 895 million to 690 million You may want to spend a few minutes getting famili
272. he start of each run None local or DPL program Microsoft Excel Go to Model Links OK Cancel Help Figure 9 14 Node Definition Links for Mkt 1 All Years Node Click OK Create Influence arcs from Mkt 1 Growth and Mkt 1 All Years to Total NPV The Mkt 1 All Years node is now fully defined and linked to Excel Note that the Excel formulas for the Mkt_1_All_ Years range now will not be used because the values are being determined in DPL and exported to Excel instead You can tell DPL to reset the Excel cells back to their original values formulas after a model run In general it is a good practice to do this If you are overwriting cells with formulas as in this case it is a particularly good idea to do this to maintain the spreadsheet with its initial functionality Yo 4 Go to Model Links Options Settings Check Reset Excel values formulas after run if it is not already checked Click OK Make sure Fast sequence evaluation is the default run type and Risk Profile Init Dec Alts and Policy Tree are selected Click Home Run Decision Analysis to run the model 2 0 Syncopation Software Chapter 9 Multidimensional Value Nodes After a few minutes DPL will return precisely the same decision analysis results that you obtained at the beginning of Section 9 1 1 Check the Session Log to compare results if you like This tutorial illustrated how series can be used in DPL s 1 dimensional va
273. high values that you specified in the Value Tornado to the Low and High outcomes as well as the value that was specified for the value node to the Nominal outcome gt Click OK to accept the probabilities and the previously specified low and high values Press ESC to deselect Unit Sales Your model should look something like Figure 2 33 61 Chapter 2 Building Decision Analysis Models Syncopation Software Product Strategy In house License Figure 2 33 Model with Sales Changed to a Discrete Chance Node Note that DPL has updated the Default Tree by adding the Unit Sales discrete chance node to it and moving the NPV get pay expression to the branches for Unit Sales In the Decision Tree DPL displays discrete chance nodes as bright green circles Note if you select the Unit Sales node in the Influence Diagram the corresponding Unit Sales node in the Decision Tree will turn light magenta indicating the node is selected in the Influence Diagram pane Conversely if you select the Unit Sales node in the Decision Tree pane the Unit Sales node in the Influence Diagram will turn light magenta 62 Syncopation Software Chapter 2 Building Decision Analysis Models Select Cost per Unit and change it to a Discrete Chance Node Note that the last node type changed to was Discrete Chance so it is now indicated as the default in the Change To icon Now you can simply click the Model Node Change to icon gt
274. hips among the different factors relevant to a decision DPL s Influence Diagram modeling environment provides a rich set of features to give you flexibility in defining these relationships In this chapter you will begin to work with those features A Decision Tree provides an intuitive map like representation of the sequence order and structure of a decision problem In DPL Influence Diagrams and Decision Trees work together to provide a complete definition of the decision problem You will work more with Decision Trees in subsequent chapters 2 1 VALUE NODES AND SPREADSHEET LINKS You will develop a decision model by adding spreadsheet linked value nodes to the Influence Diagram Each of these nodes will be linked to a cell in the Excel workbook called New Product Strategy xls which is an example file installed with the software A value node is a constant i e a number or calculated quantity i e a formula in a DPL decision model They can be linked to a cell in a spreadsheet or be local i e not linked and contain a number or formula in DPL Value nodes are represented by rounded blue rectangles in the Influence Diagram gt Open DPL DPL will start with an empty Model Window maximized on the right hand side of the screen as shown in Figure 2 1 23 Chapter 2 Building Decision Analysis Models Syncopation Software d DPL Workspace Modell oa yta Heig F wh 6 j m ll 6 t o 3 sa Eat Ch
275. how your display options are set you may notice that most of the probabilities appear as 0 The objective function value also may have fewer or more decimal places than you need to see Go to File Options and select the Outputs tab gt Change the decimal places for both General number formatting and Probabilities to 2 See Figure 11 3 309 Chapter 11 Using the Endpoint Database Syncopation Software General General number formatting Decimal places eg 123 46 Output scaling 0 display 1 0 as 1 0 Zero equivalent display 9 5e 013 as 0 Workspace Probabilities Display probabilities as percentages Decimal places for probabilities a eg 12 50 Figure 11 3 Outputs Tab of the Options Dialog Click OK The Endpoint Database report updates to reflect the change 11 1 1 Filtering the Endpoint Database You can filter the Endpoint Database Click Grid Sort amp Filter Filter You will see the Filter dialog similar to Figure 11 4 Filter Condition 1 Condition 2 Condition 3 Condition 4 Row must meet all any of the above conditions Custom gt gt gt Figure 11 4 Filter Dialog for Endpoint Database 310 Syncopation Software Chapter 11 Using the Endpoint Database gt For Condition 1 use the drop down list to select Pipeline Strategy Col 3 Select both of the strategies that include 10_BCM hold down the Ctrl key and click on the first and third strategy
276. icy display of the Policy Tree gt Make sure Policy Tree is checked gt Click Home Run Decision Analysis gt Click OK to the warning 244 Syncopation Software Chapter 8 Objective Function and Multiple Attributes DPL will warn you that the objective function and utility functions must be a linear combination of the attributes for sequence evaluation to work properly If they are not use Full Tree Enumeration as the evaluation method gt Click Yes The Policy Tree will look like Figure 8 10 Drill None 24 0 26 5 5 0 Drilling Costs Drill Yes 37 5 31 5 12 0 No 6 0 6 0 0 0 0 0 12 0 41 No 6 0 6 0 0 0 Test K 27 4 31 0 7 1 Drilling Costs Seismic Structure Dril _Yes 27 6 33 6 12 0 Core sample 27 4 31 0 7 1 Open 27 6 33 6 12 0 0 0 12 0 6 0 0 0 35 No 6 0 6 0 0 0 Drilling Costs Closed 84 2 90 2 12 0 0 0 12 0 24 No 6 0 6 0 0 0 ur p Figure 8 10 Policy Tree for the Two Attribute Model with Fat Policy Feature The expected values of both alternatives in Figure 8 10 are lower than in the original Policy Tree Figure 8 2 reflecting a penalty of 0 5 million per unit of land The Core sample Test is still the preferred alternative although the gap has narrowed The tree in Figure 8 10 illustrates DPL s Fat Policy feature At each node along the tree three values are now shown in brackets the expected value of the objective function
277. il reservoir sizes Had the expert given us the converse you would have made an arc from Seismic Structure to Amount of Oil DPL automatically performs Bayesian revision if the nodes are placed in the tree in the opposite of arc order so you can accept the data in either form 225 Chapter 7 Conditioning and Learning in Decision Models Syncopation Software The Influence Diagram is now complete You now need to update the Decision Tree gt You may wish to drag the splitter up to give you more room to work in the Decision Tree pane but make sure you can also see the Influence Diagram Right click on the Test decision node in the decision tree and select Detach from the context menu Select the branches of Seismic Structure and delete Revenues from the get pay drop down in the Model Get Pay group gt Double click on the Test decision to bring up the Node Definition dialog Select Asymmetric for the Alternative Grouping Click OK If the Wildcatter decides to conduct the Core sample test s he will know the Seismic Structure before making the decision about whether to drill Therefore the Seismic Structure chance node needs to be added to the Core Sample branch of the Test decision gt Drag the Seismic Structure chance node and drop it on the blue triangle for the Core Sample branch of the Test decision Your model should look like Figure 7 12 Drilling Costs Serene eismic Test Structure Low Drill
278. ill make one more change to the decision tree to make it even more asymmetric Suppose that management wants to know the impact on the overall expected value of the Develop decision if they were to decide that they will not launch the product if the Key competitor outcome is Breakthrough You can very easily answer this question using a simple modification to the decision tree gt Activate the Asymmetric model again gt Click in the Decision Tree Double click on the Key competitor outcome node to open the Node Definition dialog See Figure 6 32 Node Definition Key competitor outcome General Display options Data E Show name W Underline name E Border around name Links Tree Instance Ij Altemative grouping gt Symmetric a Asymmetric Mixed a Figure 6 32 Tree Instance Dialog Mixed Outcome Grouping 210 Syncopation Software Chapter 6 Asymmetric Trees Click the Mixed radio button under Alternative Grouping The Select Groups dialog appears Select both the Failure and Disappointment outcomes by using your mouse button and the Shift key Click on Group These two outcomes become the first outcome group Select the Breakthrough outcome and click on Group again The dialog should look like Figure 6 33 Select Groups Outcomes Groupe 1 Failure 1 Disappointment 2 Breakthrough Group gt gt Ungroup lt lt Heset Figure
279. imal places for probabilities 0 eg 13 Copy paste Produce simplified metafiles for ungrouping in PowerPoint OK Cancel 2 Figure 1 6 Options Dialog with Outputs Tab Selected All of the options on the General tab are at the installation level i e they are not Workspace specific The options on the Outputs and Workspace tabs are Workspace level options and can vary from Workspace to Workspace 21 Chapter 1 Introduction Syncopation Software 1 5 WHATS NEW IN DPL 8 This section is intended as a reference for users of previous versions of DPL A list of some of the major changes and new features in DPL 8 is provided below If you are a new DPL user you may skip this section DPL Standard has become DPL Direct which is a pure Excel Add in I e you do your DPL modeling directly from Excel there is no standalone DPL application DPL 8 Professional is still a standalone application With DPL Professional you also have access to the intuitive Excel side interface if you prefer to work directly from Excel The DPL 8 Direct Quick Start Guide provided with your license walks you through the essential features included in the new Excel side interface This DPL 8 Professional User Manualis a comprehensive guide to all of the features offered by the DPL 8 Professional standalone application 1 5 1 New Features in DPL 8 Professional e Microsoft Office style command ribbon user interface e Completely re worked
280. in the current model DPL has also filled in the Current setting for each variable in the table You will now enter the range for each of the value nodes shown in Figure 5 12 gt Click the blank cell in the Low column of the first value node Bad Debt Allowance Specify 0 02 for the Low value of Bad Debt Allowance Specify 0 04 for the High value of Bad Debt Allowance Continue to fill in the Value Tornado Setup table with the rest of the High and Low values shown in Table 5 1 Note that you can use the arrow keys to move around the Value Tornado Setup table GUD 153 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Gross Margin 0 30 0 40 Other Cost Growth Rate oo 0 12 RMC Lawsuit_Settlement 250 Share of Revenue at Risk Share of at Risk Revenue Lost 0 0 Trendline Revenue Growth Rate 50 05 Working Capital 100 Table 5 1 Data for Value Tornado Diagram Your completed Value Tornado Setup dialog should look like Figure 5 13 154 Syncopation Software Value Tornado Setup Modell Value Bad_Debt_Allowance Gross_Margin Other_Cost_Growth_Rate RMC_Lawsuit_ Settlement Share_of_ Revenue_at_Risk Chapter 5 Building and Analyzing Monte Carlo Simulation Models Current 0 03 0 35 01 65 01l Share_of_at_Risk_Revenue_Lost bs 0 7 Trendline_Revenue_Growth_Rate 0 0 08 Working_Capital Value Bad Debt_Allowance ka 0 035 Conditioning 120 Remove
281. ine Strategy are either Mkt 1 10 BCM or Mkt 1 20 BCM for these first few thousand endpoints Paths through the tree with these two alternatives do not include the Market 2 Political Stability Future chance event You can navigate around the Endpoint Database report as you would an Excel spreadsheet gt Press Ctrl End to scroll down to the bottom rightmost cell You can see that there are 32 112 endpoints in the database gt Press Ctrl Home to go to the top leftmost cell Note that by default the Endpoint column or column 0 is frozen so that when you scroll to the right the Endpoint column remains visible You can freeze multiple columns or no columns by entering the number of columns you d like to freeze in the Freeze column dialog via Grid Format Freeze For example if you enter 3 into the dialog and then scroll to the right you may need to make the application window smaller to scroll columns 0 1 and 2 should remain visible You can drag the vertical lines separating the columns in the first row to change their width You can also click Grid Format Columns Width to specify the width for a particular the column If you d like to specify a width for all columns in the database check the checkbox next to Apply to all in the Column Width dialog Select Grid Format Row Hide Unhide or Grid Format Column Hide Unhide and experiment with hiding and unhiding a row or column Depending on
282. ing Aad Add From tor Porta Opg Peroetual Sanal Norte beeor Figure 2 1 New Workspace with Blank Model Window Click the Model Links Add split button Note that initially the default action of this button is to add Excel Calculation Linked nodes which is what you want If you select another item from the split button s drop down list the default action changes to that So the split button allows you to execute whichever item you last executed again without having to drop down the list In the Create Calculation Linked Values from Excel dialog click Browse and select the file New Product Strategy xls It will be in the folder with DPL examples usually C Program Files x86 Syncopation DPL8 Examples 24 Syncopation Software Chapter 2 Building Decision Analysis Models Create Calculation Linked Values from Excel Spreadsheet C Program Files Ke6 Syncopation OPLeye samples ew Product Strategy xls OF Cancel Figure 2 2 Create Linked Values Dialog gt Click OK The Cell Names dialog appears The Cell Names dialog displays all named cells in the spreadsheet that are Suitable for linking to value nodes Important note DPL only displays named cells in the Cell Names dialog Before using a spreadsheet with DPL you should ensure that cells you wish to link are named in Excel Please consult your Excel documentation if you are not familiar with named cells ranges as they are essential for
283. ing Costs None d None Med Nominal Drilling _Costs lt com tame Core Sample Open High Soaking High lt Drilling_Costs Closed Figure 7 12 Decision Tree with Test Decision Added 226 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models gt Drag the Seismic Structure node on the Core sample branch down to make some room gt Drag the Test decision node the Seismic Structure node will move as well to the front of the decision tree gt Drag and drop the Drill decision node and drop it on the blue triangle for the None branch of the Test decision Your decision tree should now look like Figure 7 13 Drilling Oil Costs Low Med W Nominal 1 Drilling_Costs High Soaking High Drilling_Costs Seismic Structure None Core Sample Figure 7 13 Decision Tree with Seismic Structure Uncertainty Added On the Core sample path through the tree the Seismic Structure uncertainty is followed by the Drill decision and from there on the structure is the same as the top part of the tree Rather than repeat that sequence you ll use the Perform Subtree feature Select the Drill decision in the tree and click Model Tree Subtree on the ribbon You can also right click the node and select Perform Subtree from the context menu Place the semi transparent node over the blue triangle after the Seismic Structure uncertainty Click the left button 22 Chapter 7 Co
284. ion Software amp Cash Flow xls Figure 16 5 Workspace Manager with Item for Referenced Program A DPL program can be linked to more than one model If you have programs in your Workspace that are not linked to any model they are displayed beneath a Workspace Manager item called Unreferenced programs Programs beneath the Unreferenced programs item in the Workspace Manager are displayed with an icon that does not have a link gt Run the model with the converted spreadsheet You should notice a marked improvement in runtime gt Click the Log tab of the Workspace Window Look for the Elapsed time entry near the bottom The reduction in run time is dramatic This can be particularly useful for large models Once a linked spreadsheet model has been converted to a DPL program subsequent changes to the converted model can only be made by changing the DPL program See Chapter 15 for a discussion of DPL programs and code Usually if you plan to continue modifying the linked spreadsheet and re running the model you will not want to modify the converted DPL program Instead you can save a copy of the linked model either as a separate da file or by duplicating the model within your Workspace and revert to the saved linked model each time you wish to continue with changes to the spreadsheet Then you can re convert the spreadsheet to a DPL program each time you want to run the model with your most recent changes 432
285. ion Software Appendix A Overview of Spreadsheet Linking DPL adds the new values to your model as value nodes but does not add influence arcs You may need to add influence arcs as well as changing the node type s to decision or chance nodes See Section 2 1 for an example of this method When you use this method you will notice that DPL lets you check whether to show cells that are already linked the default is no whether to show cells with formulas the default is yes and whether to create arrays based on range size DPL assumes that since you are adding a new linked node you probably do not want to choose a cell that is already linked to another node and you probably do want to consider cells with formulas since they could become new import nodes You can check or uncheck these options if needed and the contents of the Range Names list will be updated This method allows you to create linked series and one or two dimensional arrays When creating the nodes DPL examines the spreadsheet to see what each range contains For single cell ranges that do not contain a formula DPL creates export nodes initialized with data from the spreadsheet For single cell ranges that do contain a formula DPL creates import nodes with no data For multiple cell one dimensional ranges that do not contain any formulae DPL creates export array nodes initialized with data from the spreadsheet For multiple cell one dimensional ranges that do contain a fo
286. ion from Endpoints from the list or press the keyboard shortcut for Playing Endpoints Alt F10 gt Click Ok to the warning 316 Syncopation Software Chapter 11 Using the Endpoint Database DPL takes just a few seconds to play the endpoints and produce the results Examine the Policy Tree and or the Session Log You can see that the expected value increased slightly to about 35 2 If Operating Costs are known before the Pipeline Strategy decision has to be made the gain in value is about 0 3 35 2 34 9 Note that you could have renamed and saved the original Policy Tree and Risk Profile Since you did not rename them they were over written when you played endpoints You will switch your model back to its original state But before doing so take a moment to drop down the Undo split button where you will find a list of the actions last executed In this case you will see Undo Reorder Node as the last action taken When there are multiple actions in the list you could choose to undo more than one of the previous actions in the list You can set the number of undo levels in File Options General DPL allows you to set the number of Undo levels separately for graphics windows and for text windows DPL Code Switch back to the Model window Click the Undo button on the Quick Access menu located at the top left of the application window This will undo the last action executed 11 2 2 Changing Probabilit
287. ion from Endpoints will be covered in Chapter 11 The dialog box launcher for the Run group contains a number of run options Most of the options are self explanatory A few of the options will be discussed later in the manual in the sections on Risk Profiles and Time Series Percentiles If you haven t already done so start DPL and open the workspace you saved from Chapter 2 by using File Open or pressing Ctrl O Note if you re starting here and haven t worked through the tutorial in Chapter 2 open the workspace New Product Strategy da from the Examples folder in your DPL installation folder Make sure the License vs In house prob model is the active model gt The Decision Analysis options selected in Home Run should be as shown in Figure 3 1 76 Syncopation Software Chapter 3 Decision Analysis Results F M Risk Profile W Policy Tree Levels G all e Objective Function Policy Summary _ Endpoints Decision T O aa l Analysis H Init Dec Alts O VOC C Value Correlations Run Figure 3 1 Decision Analysis options for Full Model Leave Evaluation method at Fast sequence evaluation Make sure Risk Profile is checked Leave Drop down list as Objective Function Make sure Init Dec Alts i e Initial Decision Alternatives is not checked Make sure Policy Tree is checked Make sure Number of levels is set to 6 all LUUL QQ UY Make sure Policy Summary Expected Value of Pe
288. ion policy is to perform the core sample test and then drill unless the outcome is no seismic structure See Figure 8 2 236 Syncopation Software Chapter 8 Objective Function and Multiple Attributes Drill None 26 5 Drilling Costs Drill Yes 31 5 No 6 0 41 No 6 0 Test B 31 0 Drilling Costs Seismic Structure Drill Yes 33 6 Core sample 31 0 Open 33 6 6 0 35 No 6 0 Drilling Costs Drill Yes 90 2 Closed 90 2 24 No 6 0 Figure 8 2 Policy Tree for the Single Attribute Model The expected value of proceeding with the core sample test is 31 million whereas drilling without the test has an expected value of 26 5 million Navigate back to the Wildcat model by double clicking its item in the Workspace Window or pressing Ctrl F12 The oilfield in question is in a pristine wilderness area Testing and drilling would require that some areas be cleared and that an access road be built To take this into account in your decision you ll add an attribute representing the amount of land that would need to be disturbed gt To add a value node to the model click Model Node Add value node should be the current default Place the node above Test Costs Name it Land Disturbed Click the Data tab to enter data for the new node Click the Conditioning button Check Drill and Test See Figure 8 3 GUY 237 Chapter 8 Objective Functions and Multiple Attributes Syncopation
289. iple nodes by holding down Ctrl and selecting the nodes to be copied as you did earlier when you moved multiple nodes Double click Sales 1 copy to edit it On the General tab change its name to Sales 2 Check Separate probability and value input trees as shown in Figure 3 16 Node Definition Sales 2 Values Bales _ Hide 2 Probabilities V Separate probability and value data input trees Links DPL variable name for expressions Sales_2 Outcomes Low Add Nominal Default High Inset Delete Rename F2 Default Comments Print All Comments Copy to Clipboard o ca He Figure 3 16 Node Definition General Tab for Sales 2 95 Chapter 3 Decision Analysis Results Syncopation Software Note that there are now two tabs in place of the Data tab called Values and Probabilities respectively This is convenient if you have a discrete chance node whose probabilities are conditioned by one or more nodes and whose values are conditioned by a different set of nodes or whose probabilities are conditioned but values aren t or vice versa As you will see in the Sample model you are building the probabilities for Sales 2 will be conditioned by one node while the values will be conditioned by another node Select the Values tab gt Click Conditioning The Conditioning dialog appears gt Check Price Change as shown in Figure 3 17 Conditioning Sales 2 The values for this node depend
290. ision node It is called the value of control because not only do you know which state the node is in before having to make a decision but you also get to choose which state is optimal The value of control for Sales 1 is approximately 75 This means that the expected value of the objective function would increase by 75 if you could choose the optimal outcome upfront for Sales undoubtedly High in this case and then decide on Product Strategy rather than deciding on Product Strategy and gambling on Sales 109 Chapter 3 Decision Analysis Results Syncopation Software 3 5 VALUE CORRELATIONS In this section you will examine a simple example of Value Correlations Value Correlation charts are bar charts that display the correlation coefficients between each value and event in the model and the objective function Correlation coefficients are measures of statistical association between two variables Correlation coefficients can vary between 1 0 perfect positive correlation and 1 0 perfect negative correlation Perfect positive correlation 1 0 means that as one of the two variables increases the other variable always increases proportionately Perfect negative correlation 1 0 means that as one variable increases the other variable always decreases proportionately Correlation does not always imply causation that is even if two variables are correlated it does not necessarily mean that an increase or decrease in one variable
291. isplayed in the Risk Profile Setting the number of intervals to greater than 500 will not affect how the Risk Profile charts looks since there are many fewer scenarios than intervals If you have distributions with very long tails you may wish to increase the number of intervals DPL displays the expected value EV of the objective function on the Risk Profile chart by default Displaying the expected value of the model is particularly useful when comparing multiple Risk Profiles It is also helpful to display the expected value if the Risk Profile of the model is less symmetric because it may be difficult to estimate the expected value from the risk profile If you d like to view your chart as a Histogram do the following Activate the Risk Profile Chart and check the checkbox next to Histogram in the Format Display group gt Uncheck the EV CE Lines and EV CE Values checkboxes in Format Display DPL updates the Risk Profile chart as shown in Figure 3 8 87 Chapter 3 Decision Analysis Results Syncopation Software gt Fa oO Q O fus oO Figure 3 8 Frequency Histogram Type Risk Profile Chart A frequency histogram displays the Risk Profile information in a different format The x axis still displays the objective function and the y axis still displays probability values The height of each bar in the frequency histogram Risk Profile chart indicates the probability that the objective function fall
292. isturbed and modify the other Get Pay expressions in the Decision Tree Double click on the Yes branch of the Drill decision The Get Pay tab of the Branch Definition dialog appears The dialog indicates that the model now has two attributes Profit and Land so two Get Pay expressions must be defined You don t need to contribute anything to Profit on this branch so you ll just add zero as a placeholder for that attribute 242 Syncopation Software Chapter 8 Objective Function and Multiple Attributes gt Type 0 zero in the cell for Profit it is magenta to show it is selected Click on the cell for Land Click the Variable w button and choose Land_Disturbed Click OK The dialog should look like Figure 8 8 Branch Definition Get Pay Drill Get or pay expressions Control cee CE PEPE FOU OOP OPEC OEE TOY Oe CUE DOERR ase a dbl Profit Display R ssossnnnnnn sae Land Land_Disturbed Go to Attributes Get C Pay LI YII Cancel Help Figure 8 8 Get Pay tab of the Branch Definition Dialog with Two Attributes The Get Pay expression on the Yes branch of Drill now accounts for both attributes Note with multiple attributes you can still continue to edit Get Pay expressions directly from the ribbon using the Model Get Pay Attribute drop down list and the Model Get Pay combo box as you did previously Using the ribbon method may be faster but only allows you to see the Get Pay expressio
293. ith initial model you built 2 3 RUNNING A MODEL You are at the point where you can run the model You may wish to save your Workspace before running the model Click on the Home tab Within the Home Run group there are a number of different output and run options to consider prior to a Decision Analysis run We will cover these options in more detail later For now uncheck Risk Profile and Policy Tree as these outputs would be uninteresting see Figure 2 11 TZ Risk Profile Policy Tree Levels none none C Policy Summary Endpoints Decision pes Analysis Init Dee Alts Ly WGI Value Correlations Run Figure 2 11 Decision Analysis Options in Home Run Group Note that Fast Sequence Evaluation is currently the default Evaluation method indicated by the icon in the Home Run Decision Analysis split button This is the desired evaluation method so click the Home Run Decision Analysis icon to run an analysis We will cover more on evaluation methods later gt The Model Get Pay dialog appears as shown in Figure 2 12 There is one last piece of information you must confirm for DPL which is the variable in the model that should be used as the output of the model DPL suggests NPV which is correct 35 Chapter 2 Building Decision Analysis Models Syncopation Software Model Get Pay Enter the variable or expression that you would like DFL to calculate as the final output e g Profit NPY To
294. ity Analyses Syncopation Software One a a _ _ __ _ Ca ae Ae en a L neem Rn 1 o nnn nzns nnn ue ae om T Sida DPL Product da Licence va ba heure downsteeem EA Fic tame Model Whew Guta help ye ed to i Gelete 4 Cut Th ieden Aa 4 Riak Profile Policy Tret hevels Fia r Fante Cirie Sameier tomado Time Seriei Pobre 4j Copy 2D finattepice 1 Gipicirat Punden PHY UA F Lidice infil 1000 7 Correale eeri an Ania P ta Vilut Brave Ceronsdr vngigig W mit Ore ans vou Vehet Conelaiiond Aria 160 z h j oan i Workspace Manager 4 Gi SAn Workppiece a gt Lienas ve in houdte dimanches gon Log 4 DAL heon Popranmng Language Cipy BUCO by Syiccgaiion ac Maai ice ponnng ae femre or bensa bo Sancopation Sotbhware lire AJni peweryed PC y prismi e niim wipri reipi cor US kres 1 Ge Te 7375 Datiig m US 1 Eri ol oe Protessiosal venon Perpetuai Meiras 3 50 00 Soral Murar SYPE TANAH Produd Sales Piee Perna Markeiing Saes 2 Ltt iar dakar Syrii Solaire Fe Sl ae Lier Hi Costs Coss p Low SEER In house fie Raise Nominal ff Prat J N a Nomina ah i a Nominal Hominal E License cxf Keep same Sane High d Proti lt High is Hig iy High Prom k 4 a t x Se License vt In house downn H ae Fed Help pice FE foam Te Fal Pec i Figure 14 1 Product da Workspace This model is similar in concept to the model
295. ity and determines the value of the objective function for each outcome Gamma Distribution A named distribution supported by DPL For a detailed description see On Line Help Gaussian Distribution Also called Normal Distribution A named distribution supported by DPL For a detailed description see On Line Help Geometric Distribution A named distribution supported by DPL For a detailed description see On Line Help Get Pay Expressions Expressions specified on branches of the decision tree to associate values with branches and endpoints Get expressions are added to the value function and Pay expressions are subtracted If a model has multiple attributes the get pay expression must contain expressions separated by commas for each attribute H Hide Intermediates An option when creating a model from Excel that prevents intermediate cells from being displayed in the Cell Names dialog used to select which spreadsheet cells to include in the model If intermediate are included the linked nodes will be Dummy Nodes Hyperexponential Distribution A named distribution supported by DPL For a detailed description see On Line Help Import Node Import Variable A node linked to a specific cell in a spreadsheet that retrieves the output of spreadsheet model calculations for use in DPL analysis A DPL Import Node does not contain any data in DPL Influence Arc An arc drawn from one node to another in an Influence Diagram An infl
296. ive outcome grouping for Seumetic _ C Asymmetri CS newly added nodes E Symmetric Es i 4 Ci Advanced settings Set states of events in decision tree Risk tolerance and utility functions by branch Use these settings for new models Figure 6 15 Tree Options Dialog with Tree based dependency checking This is what you want so click Cancel without making any changes Examining the arcs in the Influence Diagram one would expect NPV to depend on all four of the uncertainties but three of them are not known when NPV is calculated on the branch where FDA Approval is No The three omitted uncertainties all affect revenues and the spreadsheet knows to turn off all revenues in the disapproval case so there isn t really a problem By having this option on DPL assumes that the tree is valid and that the dependencies in the Influence Diagram don t apply to every scenario Warning To safely use this option you must make sure that the nodes that are skipped do not affect spreadsheet calculations in the scenarios where they are not known In those scenarios the values for the default state decision alternative or chance outcome will be sent to the spreadsheet For example when FDA Approval is No DPL sends the default in this case the middle or Nominal values for Market Share Market Size and Pricing to Excel before telling Excel to recalculate NPV The Excel model must be coded to ignore these values when FDA Approval is No
297. k OK gt Type 10 for From 370 Syncopation Software Chapter 13 Risk Tolerance gt Type 50 for To gt Type 5 for Step size The Run Rainbow Diagram dialog should look like Figure 13 6 Rainbow Diagram License vs In house prob Variable Variable name Risk Tolerance Select Current value S0 From 10 Number of steps 9 Step size z Evaluation method Fast sequence evaluation Change Cancel Help Figure 13 6 Run Rainbow Diagram dialog gt Click OK The rainbow diagram in Figure 13 7 shows that the optimal decision policy would change if your risk tolerance dropped to 35 The precise threshold is between 35 and 40 You could run another Rainbow Diagram with a smaller step size if you need to know more precisely the value at which the policy changes 371 Chapter 13 Risk Tolerance Syncopation Software Certain Equivalent 30 Risk Tolerance Figure 13 7 Rainbow Diagram on Risk Tolerance 13 2 ADVANCED UTILITY FUNCTIONS If you want to specify a utility function other than the standard exponential you can do so in the Risk Tolerance dialog via Model Objective amp Utility Options You need to specify both the function and its inverse using the objective function variable Utility functions cannot depend on the state of events although they can and generally should depend on or use constant values such as Risk Tolerance Figure 13 8 shows the d
298. k OK Import Endpoints From 9 WPrograrn Files 66 5yncopation OPL6 E xamples endpoint merge 2 csv Browse File Contents adpoint Database endpoint merge 2 Update Contents gM erge with existing endpoints Display report Cancel Figure 11 27 Merging an Endpoint Database DPL updates the Endpoint Database Note that the Workspace Manager does not change There is still one Imported Endpoint Database Double click on the Imported Endpoint Database to create and display the report Scroll to the bottom of the Endpoint Database Ctrl End You can see that the Endpoint Database has 32 112 endpoints Play Endpoints to see results Your model has the full merged Endpoint Database and results available The expected value is still 34 9 Close the Workspace without saving changes The tutorials in this chapter provided you with the basic steps you need to record manipulate view play import export and merge Endpoint Databases 339 Chapter 11 Using the Endpoint Database Syncopation Software 340 Syncopation Software Chapter 12 Advanced Decision Analysis Results 12 ADDITIONAL DECISION ANALYSIS RESULTS This chapter discusses additional decision analysis results and outputs available in DPL Professional including further features of Policy Trees and Policy Summaries Scenario Risk Profiles and Option Value Diagrams Chapter 3 also provides an over
299. k OK to continue 145 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Create Calculation Linked Values from Excel Spreadsheet Name C Program Files x86 SyncopationOPLE E samples E nlightity ls OK Cancel Figure 5 5 Create Values from Excel Spreadsheet Name dialog DPL will display a second dialog listing all the named cells in Enlightify1 xls You want to create value nodes for all of these cells except Discount_Rate Interest_Rate_on_WC NPV and Valuation_Date Select the cells as shown in Figure 5 6 To select or de select cells in the list you need to hold down the Ctrl key while clicking each one Range Names C Program Files x66 Syncopation DPL amp Examples Enlightify xls Creating Calculation Linked Values Location Type Count Rows Col Bad Debt Allowance Assumptions E 24 Value 1 Select All Discount Fate Assumptions C 10 Value Gross_Margin Assumptions C 30 Value Interest _Aate_on WL Assumptions 0 11 Value NPY DCFIfC 43 Formula Other Cost Growth Rate Assumptions C 34 Value SH LR SS eye 6 Assumptions C0 36 Value See fe meses Assumptions E Value Cancel Sort by location p i i Show ranges Sef 8 peas ls E Assumplonsl l 26 Value already linked Total Value Assumptions Ci7 Formula SS sate Emi Assumptions tE Value Assumptions C 12 Value Assumptions C 32 Value MW Show ranges with formulas l eo l l l b l l l l l e
300. key click on the Share of Revenue at Risk value node and the Share of at Risk Revenue Lost node Both nodes should now be magenta because they are selected Drag both nodes to the left Use this same technique or just move nodes one at a time so that the nodes in your diagram are arranged in a circle with Total Value on the right Arcs arrows in Influence Diagrams show relationships In our model Total Value depends on all the other value nodes You could add an arc from each node to Total Value but DPL has a single command to do the work Click Model Influence Arc From Formulas DPL draws arcs from each of the other eight value nodes to Total Value Your model should now look something like Figure 5 10 For information on how to create arcs individually see Section 2 2 150 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models Figure 5 10 Influence Diagram with Influence Arcs At this point you ve put enough effort into building your DPL model that you should save it Select File Save As and give your Workspace a name such as Enlightify da 151 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software 5 4 RUNNING A VALUE TORNADO DIAGRAM A Tornado Diagram is typically the first analysis done on a new model At this point your model is deterministic consists only of value nodes so the Value Tornado Diagram i
301. l Links Risk Analysis Links Eetemal Links Settings spreadsheet Name Type Calcy Init Nodes c program files 26 syncopation dplo examples enlightify Microsoft Excel 9 0 of 9 Links to Programs in the Workspace DPL Program for Data Definitions DPL Calculation Program Chan Values defined in this Choose This position is normally used for converted program can be used in spreadsheets Values defined in this program aa both the influence diagram ee can be used as calculation links or in the Remove and the decision tree oe decision tree Update Update Revert to Linked Figure 5 26 Model Links Dialog 1 0 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models Next you ll create two new DPL nodes for these new inputs in the spreadsheet gt Click Model Links Add Excel Calculation Linked should be the default gt The Range Names dialog is displayed Select Spin_off and Loss_of_Synergy hold down the Ctrl key while clicking the second item Figure 5 27 Range Names C Program Files x86 Syncopation DPL8 Examples Enlightify2 xls Creating Calculation Linked Values Name Location Type Count Rowe Col 1 1 SelectAll 1 1 Discount Fate Assumptions 0910 Value Interest Aate _on Wl Assumptions C 11 0 Value Loss of Synergy Assumptions C24 Value DCFIFE 44 Formula Spin_ott Assumptions C e Value Valuation Date Assumptions 0412 Value Cancel E Sort b
302. l beneath it Select the line Go to Chart Series Lines Markers In the Weight edit box change the number from 2 to 3 This will thicken the weight of the line Go back to the Format Display group Check the checkboxes next to X Grid Lines Y Grid Lines and Dashed Grid to format the grid lines in the background of the chart Select the line within the chart and use the Color fill bucket split button to select a new color for the line B 5 O ae je E Qo amp z oO Expected Value Figure 3 9 Formatted Risk Profile Chart 89 Chapter 3 Decision Analysis Results Syncopation Software Finally you will examine the statistics available for the Risk Profile gt Click Data Distributions Statistics You can also right click on the Risk Profile and select Statistics from the context menu The Statistics dialog appears as shown in Figure 3 10 Statistics Distribution Expected Value Distribution statistics have also been Mean 107 7 written to the log Median 105 1 Percentile 1 Mini 27 6 to Max 302 2 a0 90 Variance 6260 5 99 none Standard Deviation 79 1 ee none Skewness 0 3 r Kurtosis 23 Update Percentile Values Figure 3 10 Risk Profile Statistics The Statistics dialog gives you a standard statistical description of the distribution of your objective function including the minimum and maximum values the variance and standard deviation the higher m
303. l blank models start in default tree mode meaning DPL will automatically create then update a symmetric decision tree based on changes made in the influence diagram Starting in Section 6 1 and in subsequent sections you directly edited the default tree by introducing asymmetry to the tree causing DPL to no longer be in default tree mode Once DPL is not in default tree mode DPL will continue to add new instances to the tree with those instances being placed at the end of the longest subtree E g DPL has attached the two nodes you just added to the Influence Diagram to the Pricing node since that node previously ended the longest subtree 206 Syncopation Software Chapter 6 Asymmetric Trees If there is a Get Pay expression on the branches of the node at the end of the longest subtree and DPL places a new instance at the end of that subtree the Get Pay expression is moved from the branches of the node that was previously at the end of the subtree to the newly added node i e the PV_Market Get Pay expression was moved from the branches of the Pricing node to the Key competitor outcome node and then from there to the branches of the Launch decision You need to re order the decision tree to put the Launch decision and Key competitor outcome in the correct places But prior to re ordering the tree you will put the Get Pay expression back on the branches of the Pricing node To do so efficiently you will use DPL s cut and paste feature
304. l s Home B d Arial Ba Paste Insert Page Layout Formulas Clipboard A Data EE Appendix A Overview of Spreadsheet Linking Review General lt 0 00 00 0 Number View i a Conditional Formatting 7 H Format as Table Cash Flow with Decision xls Compatibility Mode Microsoft Excel Developer Cell Styles 7 Styles 3 Insert 7 g Delete v p Format Cells a cfs N g Z z Sot amp Find amp lt 2 Filter Select Editing Probabilities and Values for Chance Nodes Marketing Phase 1 Probabilities Values Marketing Phase 2 Probabilities Values Low 0 3 5000 0 3 8000 Nominal 0 4 8000 0 4 10000 High 0 1500 200 Marketing Phase 2 Conditional Probabilities Marketing Phase 2 Marketing Phase 1 Low Nominal High Low Nominal 0 25 0 5 0 25 High 0 1 0 3 0 61 Initialization CF 3 ila il Average 0 333333333 Count 9 Sum 3 A 100 g DEE SE ai Assumptions amp Output Ready a gt f G Figure A 11 Excel Range with 9 Probabilities for Conditioned Node To use this named range to initialize the Marketing phase 2 node you follow the same steps as before linking the first top branch of the node to the range Mkt_ph2_cond_probs The Data tab of the Node Definition dialog will look like Figure A 12 once the initialization link has been established 451 Appendix A Overview
305. l typically select several export values and at least one import value formula from a list of cells range names in your Excel spreadsheet The basic steps are as follows Click Home Add to WS Excel Linked Model Browse to locate the spreadsheet you wish to use Select the names of the ranges you wish to use DPL creates an Influence Diagram with arcs representing the relationships in your spreadsheet This provides you a starting point for your linked model See Figure A 1 and Figure A 2 below Create Model from Excel E Spreadsheet C Program Files x86 5yncopation OPLe e samples Cash Flow with Decision sle Ok Cancel Figure A 1 Create Model from Excel Dialog 438 Syncopation Software Appendix A Overview of Spreadsheet Linking Range Names C Program Files x66 Syncopation DPL8 Examples Cash Flow with Decision xls Creating Model from Excel OF Name Location Type Count Rows Col Consulting phase 1_ growth Assumptions amp Outou Value Select All Consulting _phase_2 growth Assumptions amp Outpu Value Discount_rate Assumptions amp Qutpu Value Initial consulting Assumptions amp Qutpu Value Cancel Initial license Initial training License phase _1_ growth License phase 2 growth Marketing phase 1 Marketing phase 2 Mkt phil probs Mktphl_ values Mkt_phe_cond_probs Mkt_phe_ probs Mkt_phe_ values NPY Payment_fee Pursue_Consu
306. lay any probabilities for imported endpoints You can see this is the case in the endpoints you just imported Scroll to the right to see the probability column It is blank Request a Policy Tree and then play the endpoints to check that the model results are as before with an expected value of 34 9 gt Before continuing to the last tutorial switch to the Energy Model and press F9 to clear model memory again CETA Chapter 11 Using the Endpoint Database Syncopation Software 11 5 3 Merging Two Endpoint Databases In this last tutorial you will import and merge two separate Endpoint Databases from two different CSV files You might need to do this for example if you have created a model that includes an initial decision with two alternatives and you have used other applications such as statistical or database software to calculate values for each alternative As long as the Endpoint Databases are compatible with your DPL model structure and have the required format as documented in the first several lines of comments in the CSV files you can merge the databases in DPL and use them in conjunction with your model Note that you can also merge Endpoint Databases from two or more DPL Workspace files You may wish to do this if you have a model with particularly long run times You could put the model on several different computers and record endpoints with for example an upfront decision controlled to a diff
307. le but realistic decision models This DPL 8 Professional User Manual is divided into 16 tutorial chapters plus appendices Chapters 1 through 7 cover the basic functionality of DPL These chapters are written with a new user to DPL in mind Various sections of these chapters do however assume a basic knowledge of decision analysis Excel valuation techniques and Monte Carlo simulation If you are new to DPL you should find that these chapters provide a thorough introduction to the skills you will need to use DPL 8 Professional effectively Chapters 8 through 16 cover more advanced features of DPL Professional This section of the manual assumes you have some basic experience using DPL as well as familiarity with decision and risk analysis concepts Excel and valuation techniques If you are new to DPL you should complete the tutorials in Chapters 2 and 3 and refer to other relevant parts of chapters 4 through 7 of the manual before covering the material in chapters 8 through 16 In general chapters 8 through 16 are intended to be stand alone although new DPL users may find that Chapters 8 9 and 1 are best read sequentially as discussed below The remainder of this section briefly describes what is contained within this manual The subsequent sections of this chapter describe how to get online help and technical support DPL installation and a brief tour of DPL The final section lists the major changes you ll find within the DPL fami
308. lick OK As shown in Figure 12 4 DPL filters out any branch in the Policy Tree whose rolled back expected value is negative Note that in Figure 12 4 there is no longer a branch for the No outcome of FDA Approval This is because if the FDA does not approve the drug then the expected value is negative Similarly there is no longer a branch for Key competitor outcome of Breakthrough The expected value is less than zero if the key competitor achieves a breakthrough outcome Lastly there are no branches shown for Launch equal to No in either of the scenarios where the key competitor fails or has moderate success 344 Syncopation Software Chapter 12 Advanced Decision Analysis Results Launch Market Share Failure 400 4 30 Yes 400 4 122 9 Market Share Yes 221 0 FDA Approval 2 0 Key competitor outcome Yes 177 9 Develop 2 0 Launch Moderate succ 221 0 Figure 12 4 Policy Tree Filtered on Rolled Back Expected Values Select the Market Share chance node at the end of the Moderate success branch of Key competitor outcome and the Yes branch of Launch Drop down the Policy Display Expand split button and select Expand Subtree from the list Click the Full button within the zoom controls or press Ctrl L to Zoom Full The Policy Tree should now look like Figure 12 5 The Policy Tree is difficult to read at this zoom level
309. lign or arrange nodes within the Influence Diagram You can add a block of text typically used for titles or explanatory notes within the Influence Diagram or Decision Tree and edit or delete the text via the View Layout Add Text split button Note that text blocks can be edited independent of the text associated with nodes node names and branch names You can toggle back and forth between the Certain Equivalents and Expected Value view for many outputs by clicking View Results CE The Certain Equivalents view is available only if a risk tolerance has been specified Lastly you can change the font font style and font size for various items within the Influence Diagram Decision Tree and output charts using the controls in the View Font group Data Tab Within the data tab you can export the data or statistics of a particular output via the Data Export icon Within Data Distributions you can view the Statistics for a distribution or reduce the distribution to a single chance node Lastly DPL Enterprise users can set up an ODBC Datasource for a Workspace or load the database schema information within the Data Database group Help Tab View or update your DPL version license and license information or register your DPL license via the commands within the Help Resources group View Help topics by clicking one of the three icons within the Help Help group Lastly you can launch the Examples Navigator by clicking the Help
310. lity of 0 25 Events and Values with the Same Name Although you cannot have two events with the same name or two values with the same name you can have an event and a value with the same name For example this definition decision Research Funding Low Medium High 30 ey 807 could be replaced with these two decision Research Funding Low Medium High value Research Funding given Research Funding 30 52 5 80 Values defined with events and values defined separately can be used in exactly the same way DPL can tell whether you are referring to the value or to the event by the context 404 Syncopation Software Chapter 15 DPL Programs and Code Zeros DPL also allows you to under specify the number of values If there are fewer initialization expressions than expected DPL will set the remaining values to 0 chance Fire Damage High Medium None 0 25 O5 50000 10000 DPL will automatically assign a value of 0 to the final state of this chance event Beware The fact that DPL does this can cause problems if you have incorrectly specified the values because you will not get an error message For example the following definition will not work as expected value Profit Revenues Costs Revenues Costs This definition will result in Profit being assigned the difference between Revenues and Costs only when Revenues and Costs are each in their first states All other times Profit will be assigned a v
311. located at the left end of the Excel edit bar that it says Ten_year_CF This is the name of this range See Figure 9 3 257 Chapter 9 Multidimensional Value Node Syncopation Software Ten_year_CF bd fe DCFID6 IF Proceed IF Include_Market_2 DCF D53 DCF D39 0 v A B Cc ie 21 RR 22 Other Assumptions 23 24 Annual Price Growth Market 1 OO ae 25 Annual Price Growth Market 2 a ee 26 Annual Op Costs Growth Market 1 OO A OE 27 Annual Op Costs Growth Market 2 i aa 28 29 Start year 2013s 30 value p O 055 _ __ O a 31 value q Ooo O J 32 Upfront Cost Escalation SE eae 33 Initial Marketing Spend 50 O 34 35 36 3 38 Upfront NPV 39 40 Total NPV 41 Total Annual Cash Flow Time Series 43 44 v 4 4 gt Assumptions DCF Market Trend Daa 2 i eb gt i Ready 23 Average S 1 1 Count 10 Sum 11 3 O 100 g Figure 9 3 Excel Cash Flow Range Switch back to DPL Activate the model double click on its item in the Workspace Manager or Ctrl F12 gt To add a new value node to the model click Model Links Add Excel calculation linked should be default add link type see Table 1 2 If it isn t use the drop down to select it In the Range Names dialog select Ten_year_CF and check the Create arrays based on range size checkbox Figure 9 4 258 Syncopation Software Chapter 9 Multidimensional Value Nodes Range Names Multidimensional V
312. ls and situations including probabilistic conditioning Bayesian updating and learning models Chapter 8 demonstrates how to create models with multiple attributes and use multiple attributes in an objective function as well as how to use constraint functions Multiple attributes are also used to track various metrics even if they aren t incorporated into the objective function This use of multiple attributes is discussed more in chapters 8 and 9 Chapter 9 demonstrates how to use multidimensional value nodes in the Influence Diagram One dimensional arrays or series and two dimensional arrays can be modeled with value nodes in the Influence Diagram These non scalar value nodes can be used to represent items such as sales over several time periods one dimension or over both market segment and time period two dimensions They can also be used to represent free cash flows over a time horizon or other metrics that might be of interest Multidimensional value nodes have several advantages They significantly reduce the work required to build an Influence Diagram and set up 2 Syncopation Software Chapter 1 Introduction spreadsheet links Multidimensional value nodes result in a much simpler easier to read Influence Diagram And lastly they are faster than multiple scalar value nodes when linked to an Excel spreadsheet Chapter 1 contains a tutorial on using DPL s time series percentiles feature including the use of multidimensional va
313. lter you just created to become a custom filter Click Grid Sort amp Filter Filter 311 Chapter 11 Using the Endpoint Database Syncopation Software Press the Custom button If you have defined a filter using the controls at the top of the dialog as you have just done DPL will translate that filter into a custom filter and show the logic in the Custom filter edit box as shown in Figure 11 6 Filter Condition 2 Condition 3 Condition 4 kd none none none Row must meet all any of the above conditions Custom filter Col_2 0 Col_ 2 amp amp Col_7 0 Figure 11 6 Custom Filter Logic In custom filters the columns i e DPL events attributes etc are denoted Col_0 Col_1 etc according to their order in the report The states of each column are integers numbered starting at zero The filter operators are as shown in Table 11 1 below In particular the filter you created uses the operators equal to logical OR and amp amp logical AND 312 Syncopation Software Chapter 11 Syncopation Software Chapter 11 Using the Endpoint Database the Endpoint Database logical AND logical OR Equal to Not equal to Less than Greater than Less than or equal to Greater than or equal to logical NOT Table 11 1 Filter Operator Definitions Place your cursor in the Custom filter edit box so that it is between the last two right parentheses Type
314. lting TOTAL _COST_OF_SALES TOTAL OVERHEADS Tramning_phase_1_ growth Training_phase_ _ growth Wages_phate_1 Wages phase Wages salaries Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Initialization 09 E 7 Initialization C S E 5 Initialization 091 r Initialization 041 0 4 Initialization 091 1 4 Assumptions amp Qutpu Assumptions amp Qutpu Assumptions amp Qutpu CFO 14 0A CFID 20 0 20 Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu Assumptions amp Outpu CFO 1 r07 Value Value Value Value Value Value Value Value Value Value Value Formula Value Value Formula Formula Value Value Value Value Formula ee ee ee ee ee ee ee Ce a a A A G d a ee a a I I 5 St l l l l l l l l l l a i tt u u i st i st M IM IM IM st _ Prefix node name a l o o l a a a A a A A A ee a e a a a I I i 5 x Sort by location name with sheet Figure A 2 Range Names Dialog for Creating Model from Excel When you use this method you will see that DPL includes all of the named cells in the Range Names dialog so that you can choose to use any or all of the cells in your linked model Also see Section 2 2 of this manual for an example of Method 1 A 2 2 Method 2 Adding Calculation linked Nodes to an Existing
315. lue nodes to give you more control and more power in building models In some situations you might want to develop more complex series using expressions that refer to other value nodes in your DPL model or you might not want to have to switch to Excel every time you want to change an assumption or a formula In this example you needed to create the node and link it via the Node Definition dialog because you wanted to create an export node and the range you were linking to in Excel contained formulas Had you used Model Links Add the Range Names dialog would have assumed you wanted to create an import node 9 2 CREATING AND LINKING A Two DIMENSIONAL VALUE NODE In the sections above you modified a DPL model to include two new value nodes a one dimensional import node and a one dimensional export node In this section you will define a two dimensional array that is an export node and link it to Excel The process for defining two dimensional arrays is very similar to the steps you used to define the one dimensional arrays Either continue with the example you saved in Section 9 1 1 above or if you closed them re open Multidimensional Value da and Multidimensional Value xls In Excel select the Assumptions sheet Press F5 the Go To button in Excel Scroll down to the Yr_By_Yr_Growth range and select it See Figure 9 15 2 1 Chapter 9 Multidimensional Value Node Syncopation Software Upfron t_Cost_Escalation u pre
316. lue nodes with this feature Because time series percentiles use multiple attributes and can be used with multidimensional value nodes for greater efficiency you may wish to review Chapters 8 and 9 before completing the tutorial in Chapter 1 Chapter 11 is a tutorial on using the Endpoint Database features in DPL The Endpoint Database can be recorded saved and viewed A recorded Endpoint Database can be played to very quickly produce decision analysis results even if the model has been modified in certain ways The Endpoint Database capability can significantly reduce the time required to do multiple runs of large models Chapter 12 covers DPL s more advanced results such as Policy Trees Scenario Risk Profiles Policy Summaries and Option Value Charts Chapter 13 shows you how to use DPL to model attitudes toward risk and make decisions based on criteria beyond expected value Utility functions are introduced in this chapter Chapter 14 covers advanced sensitivity analysis features in DPL including two way Rainbow Diagrams and Event Tornado Diagrams Chapter 14 also contains guidance about when to use each type of sensitivity analysis Chapters 15 and 16 describe and demonstrate how DPL programs and DPL code are created and how spreadsheets linked to DPL can be converted to DPL code for significant improvements in model runtime These two chapters are intended mainly for more advanced applications that require l
317. lue will be recalculated by Excel throughout Paste Link the run Initialization links These links are for connecting to a spreadsheet or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at the start of each run the spreadsheet e g NPV None local or DPL program Microsoft Excel Import nodes dont need to be initialized Figure 6 18 Node Definition dialog Links Tab gt Click OK A prompt will appear asking if you would like to replace the old node name NPV with the new node name PV Market in node data and get pay expression in the decision tree See Figure 6 19 The node name has changed Would you like to replace all occurences of the old name with the new name in node data and get pay expressions Don t ask again can be changed in File Options General Yes Ho Figure 6 19 Node Name Change Prompt gt Click Yes Note that if you d like DPL to automatically find and replace old names with new in the future then you should check the Don t ask again checkbox If you d like to be prompted each time leave the checkbox unchecked These options can be changed in File Options General in the After Node Name Change section 199 Chapter 6 Asymmetric Trees Syncopation Software DPL changes the Get Pay expression for Pricing and FDA Approval to PV Market as shown in
318. ly complete However the Default Tree is not yet correct The model you are developing is a two time period model You have an upfront decision to produce the product In house or License it You then have a downstream decision to change the price or not The downstream decision occurs after you have observed sales in the first period To reflect this you need to have the Sales 1 discrete chance node appear before to the left of Price Change in the tree As shown in Figure 3 23 Price Change is currently before Sales in the Default Tree You will now add some information to the model so that DPL will correct this Specifically you will add an influence arc from Sales 1 to Price Change As mentioned in Chapter 2 an influence arc can represent dependency conditioning and or timing In this case the influence arc from Sales to Price Change is a timing arc Price Change does not depend on and is not conditioned by Sales 1 The arc tells DPL that Sales occurs before Price Change 103 Chapter 3 Decision Analysis Results Syncopation Software gt Click Model Influence Arc Add see Table 1 3 Create an arc from Sales 1 to Price Change click on Sales then on Price Change Or use one of the methods discussed earlier to create the arc hold down Shift click predecessor then click Successor or hold down Ctrl click predecessor click successor then click Model Influence Arc Add Note that DPL has moved Sales 1 ahead of Price Chan
319. ly of products Chapter 1 Introduction Syncopation Software This manual focuses on DPL s graphical model development environment It is intended to be read while working with DPL The chapters in this manual contain tutorials that will give you the essential tools needed to develop real world decision problems in DPL Chapters 2 and 3 contain a tutorial focused on building a model starting from a simple spreadsheet representing the business plan for a new product These two chapters cover a broad range of modeling topics and cover all the major analytical outputs so they re a good place to start if you re new to DPL Chapter 4 contains a tutorial on strategy tables which allow you to aggregate several up front initial decisions into a logical manageable set of strategies to be analyzed Chapter 5 contains a tutorial focused on building and analyzing a Monte Carlo simulation risk analysis model in DPL based on an Excel cash flow model The rest of the tutorials in the manual cover more advanced topics and assume that you are already familiar with DPL or have read one or more of the earlier chapters in this guide Chapter 6 demonstrates how to build and run asymmetric decision trees Although simple symmetric trees are useful in some situations many decision problems are best modeled and communicated using asymmetric tree structures Chapter 7 is an introduction to using DPL s modeling environment for more complex decision mode
320. lysis Models Product Strategy In house License 1 Figure 2 15 Model with Product Strategy Decision In the Decision Tree pane DPL has begun to build a decision tree for you This decision tree is called the Default Tree Up to this point your model had no decision nodes in it and no chance nodes which we will cover soon so there was nothing for DPL to put in the Default Tree If you do not modify the decision tree yourself DPL will continue to build a Default Tree for you If you build your model entirely from the Influence Diagram and find once it is completed that the Default Tree has the structure you want then there is no need to change the Default Tree Once you modify the Default Tree you have taken control of the tree DPL will no longer build a Default Tree for you DPL will continue to add nodes to the Decision Tree when you add nodes to the Influence Diagram but you may need to further modify the Decision Tree for it to be correct for your purposes You can always ask DPL to create the Default Tree from the current Influence Diagram by clicking Model Tree Default Tree see Table 1 5 39 Chapter 2 Building Decision Analysis Models Syncopation Software As in the Influence Diagram DPL uses yellow squares to represent decisions in the Decision Tree DPL uses blue triangles to indicate endpoints sometimes called leaf nodes in the Decision Tree In order to incorporate the new decision into your model
321. me Per Unit Costs PerUnit_Costs High Select Current value 0 9 From Humber of steps 45 Step size Second Variable y axis Varable name Select Current value Mone From Humber of steps 45 Step size Evaluation method Fast sequence evaluation OK Cancel Figure 14 5 Two Way Rainbow Diagram After Selecting First Variable Y Gee Fes In the From edit box in the First Variable x axis section type 0 8 In the To edit box type 1 05 In the Step Size edit box type 025 Click the Select button in the Second Variable y axis section The Select Value dialog appears again Select License_Fee in the Value for sensitivity drop down list Click OK In the From edit box in the Second Variable y axis section type 0 7 In the To edit box type 0 9 In the Step Size edit box type 025 379 Chapter 14 Sensitivity Analyses Syncopation Software The Run Two Way Rainbow Diagram dialog should look like Figure 14 6 Two Way Rainbow Diagram License vs In house downstream First Variable axie Variable name Per Unit_Costs Per Unit_Costs High Select Curent value 0 9 From 8 Number of steps 11 Step size 25 Second Variable y axis Variable name License Fee Select Curent value ig 75 From at Number of steps 49 Step size 02 Evaluation method Fast sequence evaluation Figure 14 6 Run Two Way Rainbow Dialog after Specifyin
322. ments are strings of characters that are ignored by the DPL compiler They allow you to insert explanatory notes in a DPL program to organize it or to help other people understand it You can create comments that occupy several lines but you cannot nest one inside another Comments can appear anywhere a blank or tab is allowed There are two types of comments in DPL The first type is bracketed by two pairs of characters and This type of comment can occupy more than one line The first occurrence of ends the comment This type of comment may be used in the data entry fields of the Model window as well as in programs and command statements The second type of comment begins with a pair of forward slash characters and ends at the end of the line By definition this type of comment cannot occupy more than one line 393 Chapter 15 DPL Programs and Code Syncopation Software 15 2 4 Spelling Case Sensitivity It is very important to pay attention to spelling in a DPL program DPL is case sensitive which means that Oil_ Price and oil_price are not equivalent All DPL reserved words are lower case decision chance and value are all reserved words A reserved word cannot be used as an identifier User defined identifiers can be in lower or upper case Refer to the online Help for more information on identifiers and reserved words 15 3 GONVERTING A MODEL TO A PROGRAM The easiest way to familiarize yourself with DPL programs
323. mparison chart to make the window active Initially the chart will display the Expected Value risk profile dataset from the last simulation run You want it to display all three of the saved risk profiles gt Click Chart Series Data Modify from the menu The Modify Data dialog will be displayed gt In the drop down list for Series 1 select 100 Samples Select 1000 Samples and 10000 Samples for Series 2 and Series 3 respectively See Figure 5 22 165 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Data Display risk profile data from this model onby A a 4 i 4 Figure 5 22 Format Diagram Dialog Series tab Click OK See Figure 5 23 166 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models p S UJ Ww u A Le gt 5 4 A 45 QO o a 4 I 2 5 2 I 1 5 1 j l 0 5 H E mo Yl 800 600 400 200 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 Figure 5 23 Sample Size Comparison gt Uncheck Format Display Histogram to make it easier to compare the Risk Profile Datasets The chart window displays all three of the risk profiles in cumulative form Figure 5 24 You will also add a legend to indicate which color goes with which graph gt Check the checkbox next to Show within the Chart Format Legend group Try out different placements of the l
324. mples down paths in the tree Useful to reduce runtime when an exact expected value is not required A E Table 3 1 Evaluation Method Types 73 Chapter 3 Decision Analysis Results Syncopation Software Risk Profile Objedive Fmdion This section controls the probability distribution related outputs that DPL will provide for a Init Dec Alts model run Risk Profile Creates a graph showing the full range of possible outcomes from the model The format of the graph can be changed after the model run Init Dec Alts Initial Decision Creates risk profile graphs for each of Alternatives the alternatives of the decision at the head of the tree Table 3 2 Distribution Output Options 74 Syncopation Software Control E Policy Tree Levels none al Policy Summary Endpoints Run Policy Tree Policy Summary Expected Value of Perfect Information Control Number of levels Endpoints Chapter 3 Decision Analysis Results bi This section controls the decision policy related l outputs that DPL will provide VOK Value Correlations for a model run Creates an output tree that displays the optimal alternative for each decision the probability of each outcome for each discrete chance node get pay values and rolled back expected values at each point in the tree Creates a view that summarizes the information in the Policy Tree for each decision and chance node Crea
325. my Node A node that represents an intermediate calculation in a model It contains no data and is ignored during analysis The graphical symbol for a dummy node is the same as a value node a blue rounded rectangle in the Influence Diagram Dummy nodes are included for clarity or communication only Dummy nodes usually represent intermediate calculations in a linked spreadsheet Dummy nodes do not appear in decision trees Dynamic Data Exchange A Microsoft Windows communications protocol that allows Windows applications to communicate with each other DPL optionally uses DDE to communicate with Excel See also OLE Automation Dynamic Link Library A file containing functions instructions or programs for use by DPL during analysis Endpoint The terminal point of each path through a decision tree The number of endpoints or leaf nodes of a decision tree equals the number of paths through the tree Endpoints are represented as blue triangles 1 Each blue triangle in a Policy Tree is an endpoint 2 In the Decision Tree because the decision tree is schematic blue triangles can represent more than one endpoint In the Decision Tree an endpoint triangle serves as the point of connection when you add a node to the tree Endpoint Database The complete set of endpoints that result from running a decision model The Risk Profile for the best alternative of the initial decision of a model is created from the subset of endpoints in the E
326. n DPL runs your model with the new assumptions in Excel and overwrites the Endpoint Database with a new one As you can see from the Policy Tree the model results are dramatically different The expected value is now over 100 From the Policy Summary you can tell that the optimal policy has changed as well Most of the time the optimal pipeline strategy is to pursue Market 1 only with a pipeline capacity of 20 BCM You learn the new assumptions are incorrect they were too optimistic You need to immediately produce a Value of Information and Control chart for the model using your original assumptions Rather than going back to Excel removing the changes and re running the model you can generate this in a few clicks by simply reconnecting and playing the Endpoint Database you saved earlier Activate double click on the Saved Endpoints item in the Workspace Manager gt Click Grid Endpoints Reconnect DPL will ask if you wish to save the Endpoint Database currently connect to the model the one you created with the optimistic assumptions as shown in Figure 11 20 Click No j gt This will clear the current endpoint database Would you like to save it EP first Figure 11 20 Save Endpoint Database Prompt DPL reconnects the Saved Endpoints database and gives it the default name Endpoint Database as shown in Figure 11 21 Note that the connected endpoints are not yet displayed in a report
327. n Cost per Unit General Conditioning Print Page Setup Copy FullScreen Wheel Reorder QARA Distribution type discrete M Links Probability gt m Ss Value 0 8 Vv Cost Nominal In house per Unit Eoo 0 8 O High Product Strategy Cost Nominal per Unit License OK cane He Figure 2 35 Node Definition Dialog Data Tab for Cost per Unit gt Type 0 6 in the Probability edit box for the Nominal branch of the In house alternative Arrow down twice to select High Type 0 4 in the Probability edit box for High Because both outcomes have value zero the probabilities under the License branch don t affect the results but they need to be filled in gt Arrow down twice to move to the probability for Nominal under the License branch Type 0 5 in the Probability edit box for Nominal Arrow down twice to select High Type 0 5 in the Probability edit box for High Click OK to close the Node Definition dialog Change the Node Type of Marketing Costs to a discrete chance node Model Node Change To Accept the default outcomes and select the Data tab Accept the default probabilities and previously specified values Click OK to close the Node Definition dialog Repeat the above steps for Price GUC HULU 64 Syncopation Software Chapter 2 Building Decision Analysis Models Note If you don t review the data on the Data ta
328. n Analysis and generate a Risk Profile If you d like to change these settings prior to a Decision Analysis run you can access the Run Options dialog via Home Run Options Also see Section 3 2 for information on Risk Profiles and the Number of intervals setting Make sure the Number of intervals is 500 Click OK twice to run a Time Series Percentiles chart Note adding attributes to a model does have a performance impact After a minute or two DPL produces the Time Series Percentiles Chart shown in Figure 10 12 290 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics 90th Percentile 10 50th Percentile 10th Percentile 15 Expected Value 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 Figure 10 12 Time Series Percentiles Chart DPL displays the 10th 50th and 90th percentiles or the percentiles percentile ranges you specified plus the expected value of the time series variable for each time period In this case the time series variable is annual cash flow so Figure 10 12 displays the three percentiles and expected value of cash flow in each of the time periods The 10th percentile is the value such that there is a ten percent chance that the time series variable is less than that amount The 90th percentile is the value such that there is a ten percent chance that the time series variable is greater than that amount The 50th percentile is the value such that the
329. n Models Drilling Costs Revenues Med Nominal 1 Drilling_Costs Revenues High Soaking High Drilling_Costs Revenues Seismic Structure None Figure 7 14 Decision Tree with Test Costs Get Pay Expression Added gt In the Home Run group check Risk Profile and Initial Decisions Alternative gt Check Policy Tree gt Click Home Run Decision Analysis gt Click OK to the warning The Policy Tree in Figure 7 15 shows that the Core sample test is the optimal alternative Even including the 6 million cost of the test the Core sample alternative has an expected value of 31 million 4 5 million more than drilling without a test 229 Chapter 7 Conditioning and Learning in Decision Models Test 31 0 Seismic Structure Core Sample 31 0 6 0 Drill None 6 0 41 Drill Open 33 6 35 Drill Closed 90 2 24 mW Figure 7 15 Policy Tree Syncopation Software Yes No Yes No Yes Drilling Costs 31 5 6 0 Drilling Costs 33 6 6 0 Drilling Costs 90 2 6 0 When there is no seismic structure the optimal policy is not to drill since doing so has an expected value of 31 5 million The imperfect information about Seismic Structure is valuable because it changes the Drill decision in that scenario gt Navigate to the Risk Profile Chart labeled Initial Decision Alternatives in the Workspace Manager See Figure 7 1
330. n for a single attribute at a time The Get Pay tab of the Branch definition dialog allows you to see the expressions for all the attributes simultaneously You also need to change the other Get Pay expressions this time adding a 0 placeholder for the Land attribute instead of the Profit attribute 243 Chapter 8 Objective Functions and Multiple Attributes Syncopation Software Double click on the branches of Drilling Costs gt Type 0 zero in the cell for Land gt Click OK Note When you edit a Get Pay expression if you leave a cell blank for an attribute DPL will fill in a zero when you click OK to close the Get Pay Definition dialog Try this now Double click on the branches of Oil Price gt Click OK Note that zero is filled in for Land in the Get Pay in the Decision Tree gt Repeat the above process for the Get Pay on the Core Sample branch of the Test decision You have created a two attribute model with an objective function that linearly combines the attributes Your tree should look like Figure 8 9 Drilling Costs I Revenues 0 M WwW Nominal 1 0 Land_Disturbed Drilling_Costs 0 Revenues 0 High Soaking High Drilling _Costs 0 Revenues 0 Seismic Structure No Core sample Test_Costs 0 Figure 8 9 Decision Tree for Model with Two Attributes You are now ready to run your model Remember with the Gather attribute expected values checkbox checked DPL will produce a Fat Pol
331. n the other four attributes with elements of the column vector which are denoted Invest_5x1 0 0 Invest_5x1 1 0 etc Suppose you have defined a 2x5 two dimensional array called Invest_2x5 As with the one dimensional arrays you would select Invest_2x5 for Attribute2 and use the Fill Down button In this case DPL will ask you whether to fill in with the first row or the first column of the array 303 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software e You can also start filling in from an element of an array other than the first element To do this enter for example Invest_1x10 5 for the first attribute you wish to fill in and again use the Fill Down button to fill in the subsequent array elements In this example DPL would fill in array elements 5 through 9 for the five attributes e The Fill Down button works with series value nodes as well as array value nodes You may wish to experiment on your own to fully see how the Fill Down button in Get Pay expressions works 304 Syncopation Software Chapter 11 Using the Endpoint Database 11 USING THE ENDPOINT DATABASE When DPL runs a Decision Analysis it creates a variety of results and charts that help you understand interpret and display insights from your model Underlying these results is an Endpoint Database which DPL can also store or record from a model run An Endpoint Database can be envisioned as a simple tabul
332. na__ Cept Adua P Actual PL Market Forecaste Market Operstin Appraise Probabili Ot n F Mraiei w a eT J Ss ree E YE Lew TTE BCM bew Low Low table Lew Lew Lew m nopan 2 ii 212 es low Miti No BEMI Low Low Low Stable Low Low Nominal 3 Ven Low Mitt Me RER IO Liw Lim Liu Stable Lows Lows high 44 Vn Lowa Wiel Pie EEM 10 Liw Law Low Stable Low Mormmaal Litre 365 Ver Low Pii I Ho ECM l0 Los Low Low Stable Low Momma hicerinal 6 6 Wea Low Mkt M DERIO Low Low Low Stable Low Meme High Tory Yri Lowe Bikr I Mia BCM J0 Lw Litres Low Stable Lsi High L ns 5 B ve Low Mit Abe Echi lg Lpa Lyw Low ttable Low High Maning 9 g Ver Low Pikti Pe DEMIO Low Low Low Stable Low High High 10 10 Vel Low Mkt I Pia RCH JO Lr Liu Liw Stable Hormel Lives Lira H Y Lew Mii I M BLO Lim Liew Low Stable Morterkal Love Plertieral iz 17 Ye Los Wiii Ho WMI Low Low Low Ttable Horanal Low High FEE Wee Lida Kit To Hic BOM Ls Lira Liw Stable Harrarcal Pisana L ni la 14 Ye Lonia Miki I hia ECM 10 Liw Low Low Stable Horrersil Mertnal Mecrunal 4 a i 15 153 We Law hat S Ao EMG Lra Losy Low Stable Honenal Homunal high 16 15 Ve Low Wid I He DERIO Lis L w Low Stable Marrenal High L Seyon Leg unpnaa ine i ir Vr Lows Miil ha REM JO Lir Laver Lem Seable Norraral High Hernimal GS birer 1 B68 TSG 2375 ri a ais i P pia Duiske the WIE a1 SE TH 1500 14 18 Ye Lew Piit I Ale BTH lG Low Lery Low Stable Homai High High 19 13 Ve Low Piti He BOML1d Low Low Low Stabl
333. nalysis Models Insert Data Review View DFL Arial 2 Currency A galinet AT Ea E o P p styl g Delete tyles sot Find amp 8 08 felFormat 27 Filter Select Number F Cells Editing Value Drivers and Metrics Discount Rate Base Date Revenue per Unit Unit Sales Cost per Unit Marketing Costs Drivers Calculations Ey 2 3 4 5 6 7 8 9 10 11 12 16 a7 18 19 20 22 23 25 27 28 Figure 2 8 Excel with Linked Cell Selected gt Switch back to DPL Click OK to close the Node Definition dialog 31 Chapter 2 Building Decision Analysis Models Syncopation Software At this point you may want to name your DPL model and save your Workspace file GU Select the item for the model in the Workspace Manager DPL gave it the default name of Model1 Press F2 to edit the name Type License vs In house or any other name you d like Press Enter to save the edit Save the Workspace and give it a name of your choice using File Save 2 2 ARCS You will now add some influence arcs to the decision model Influence arcs indicate one or more of three things to DPL 1 dependency 2 conditioning and or 3 timing Note see the Glossary of DPL and DA Terms for definitions of dependency and conditioning Influence arcs are useful because they visually indicate which nodes depend on o
334. nce Diagram files inf from DPL 3 x into a DPL 8 Workspace using File Import You can also import DPL Program files dpl in the same manner You can back save a DPL 8 Workspace file to either a DPL 4 0 5 0 6 0 or 7 Project Workspace file using File Save As Some information for new features in DPL 8 may be lost 453 Appendix B System Requirements amp Compatibility with Older Releases Syncopation Software 456 Syncopation Software Appendix C Keyboard Shortcuts C KEYBOARD SHORTCUTS Action Ss Keyboard Shortcut Ctrl Y Ctrl T Clear Memory delete unsaved Endpoint F9 Database Policy Tree Risk Profile etc Clear All Output associated with Model Ctrl Shift F9 Ctrl C Ctrl Insert Ctrl X Shift Delete Delete Ctrl F Ctrl F12 F5 Help F1 Ctrl N Ctrl O Ctrl V Shift Insert Alt F10 Ctrl P Ctrl R F3 Ctri H Run Decision Analysis using most recent F10 evaluation method Monte Carlo simulation for continuous models and produce outputs as specified on the ribbon p b requested portfolio outputs F8 Ctrl S Ctrl A Ctrl G pane Toggle between Graphics Text Mode in Policy Tab Tree and Policy Summary Alt lt Alt F12 Alt Enter Shift F12 Ctrl L Ctrl Shift R Ctrl gt 45 Appendix C Keyboard Shortcuts Syncopation Software Ctrl lt Ctrl E 458 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms D GLOSSARY OF DPL AND DECISION ANALYSIS TE
335. ncel to close the dialog without making changes You ll now add the learning node associated with the core sample test gt Add a new Discrete Chance Node to the Influence Diagram Place it below Amount of Oil gt Name the node Seismic Structure Name its outcomes None Open and Closed gt Click OK to close the Node Definition dialog Your model should look like Figure 7 9 222 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models Figure 7 9 Influence Diagram with Seismic Structure uncertainty You know from geological data the relationship between Seismic Structure and Amount of Oil You need to add an arc from Amount of Oil to Seismic Structure gt Hold down shift and click on Amount of Oil The cursor changes to the end arc cursor gt Click on Seismic Structure The new arc will initially have a black arrowhead Note you can also add influence arcs by clicking Model Influence Arc Add see Table 1 3 clicking on the predecessor node and then clicking on the successor node gt Double click on the arrowhead to change the type of the arc DPL displays the Chance Event Influence Type dialog Figure 7 10 gt Check Different probabilities for each conditioning event state Make sure Different values for each conditioning event state is not checked 223 Chapter 7 Conditioning and Learning in Decision Models Syncopation Software Chanc
336. ncy value OperatingCost Mode 20 50 chance SystemFailure 0 01 chance CoreMelt SystemFailure Mode Oo be Prr 001 B chance Warning Clear Ambiguous None SystemFailure EOT N r E ON T NA fF sequence attributes 2 objective 1 MeltCost 2 gamble on Warning then decide on Mode and pay prob CoreMelt y OperatingCost DPL compiles the expression for the objective function before it processes the Decision Tree This means that the expression cannot depend on the states of any events If the expression contains values they cannot depend on the states of events Also the expression cannot contain the state function which returns a number equal to the state of an event When there is only one attribute DPL s default objective function is equal to that attribute The default objective function is the sum of the attributes which is equal to the first attribute when there is only one However you can specify an expression for the objective function even where there is only one attribute This allows you to multiply all the gains and losses by a scale factor before DPL uses them in an analysis It also lets you use any expression to transform the sum of the gains and losses before applying a utility function if any and determining the optimal policy The phrases attributes objective and constraint are optional sequence 2 1 Cost1 2 Cost2 You can also specify a constraint function sequence
337. nditioning and Learning in Decision Models Syncopation Software DPL labels the Drill decision node with an a This is called a Perform Target DPL also changes the blue triangle after the seismic structure uncertainty to be a yellow decision node and labels it a This is called a Perform Reference Perform Subtrees can also be headed by a chance node i e the Perform Target can be a chance node If the Perform Target is a chance node then the Perform Reference is drawn as a chance node Multiple Perform Subtrees are allowed in a DPL Decision Tree Note you can also create a Perform Subtree by Ctrl Shift dragging from the Perform Target to the Perform Reference To do this for the example above you would hold down both Ctrl and Shift and drag the Drill decision to the blue triangle after Seismic Structure Finally you need to add a Get Pay expression to the tree for Test Costs and Revenues Select the Test Costs node in the Influence Diagram pane Drag it down to the Decision Tree pane The cursor changes to the add oP get pay cursor get gt Drop it on the Core Sample branch of the Test decision gt Use the same process as above to add Revenues as a get pay for Oil Price Note as an alternative to the above steps you could have used the Model Get Pay drop down to specify the get pay Your decision tree should now look like Figure 7 14 228 Syncopation Software Chapter 7 Conditioning and Learning in Decisio
338. ndpoint Database that are part of the optimal policy Certain additional outputs can be created from the Endpoint Database without the need to re run the entire model Endpoint Database Displaying The process of displaying a recorded Endpoint Database in a spreadsheet like tabular report Displaying of the Endpoint Database can take a significant amount of memory and may not be possible with very large models Occasionally with large models there is enough computer memory available to record but not to display the Endpoint Database 465 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software Endpoint Database Recording The process of saving the Endpoint Database from a model run in order to re use it later Recording of the Endpoint Database can take a significant amount of memory and may not be possible on some computers with very large models Recording the Endpoint Database is most appropriate and useful with models that have complex value models and therefore have relatively long runtimes per endpoint calculation Erlang Distribution A named distribution supported by DPL For a detailed description see On Line Help Event A node with a set of possible states outcomes or alternatives In DPL decision nodes and chance nodes are events Event Tornado Diagram A sensitivity analysis that compares the range of uncertainty of each chance event in a model Expected Value EV
339. ne path The expected value is 23 The dont prefix only applies to the lottery it precedes It does not apply to other occurrences of the same lottery elsewhere in the sequence section 15 5 7 Getting or Paying the Value of an Expression You can associate a gain or loss with each branch event state in a Decision Tree with a get pay expression and get 20000 and pay sum Cost1 Cost2 Cost3 and receive Revenues Costs and lose Damages If you tell DPL to keep track of several attributes throughout a decision problem you must provide one expression for each attribute whenever you specify a gain or loss Multiple attributes are described later in this section The number of expressions following the keywords get receive pay Or lose must equal the number of attributes specified in the program If more than one expression follows one of these keywords the expressions are separated by commas and get Cash Flow Cash Flow2 15 5 8 Setting the State of an Event At any point in the sequence of events DPL allows you to set the state of an event with the keyword set If you set the state outcome of a chance event that outcome effectively has a probability of 1 0 at that point in the Decision Tree regardless of the initial probabilities contained in the event s definition If you set the state alternative of a decision that alternative is selected regardless of the expected values associated with other alternatives
340. ng Get Pay expression on the branches of a node in the decision tree DPL will add that value to the existing Get Pay expression e g the Get Pay expression Valuel will become Value1 Value2 Your tree should now look like Figure 6 22 you may need to Zoom Full to see the whole tree Market _ Ee Pricing Approval Low PV_Market Low Low oO Develop PV_Market L Yes Nominal Nominal Nominal lt f Yes PV_Launch PV_Market PV_Development S No High High High C No g d Figure 6 22 Decision Tree with Three Get Pay Expressions Select Home Run Decision Analysis or press F10 The results 75 million for Yes and zero for No are as before gt Shift double click on the FDA Approval node in the Policy Tree to expand and zoom the tree See Figure 6 23 DPL displays the values for get pay expressions below the node branch on which they occur in the Policy Tree In Figure 6 23 you can now see the values for development cost 174 million below the Yes branch of Develop and launch costs 123 million below the Yes branch of FDA Approval 201 Chapter 6 Asymmetric Trees Syncopation Software Market Share FDA Approval Yes 323 7 74 8 50 122 9 174 0 Develop 74 8 Figure 6 23 Policy Tree With NPV split into its components you are adhering to all the arcs in the Influence Diagram You no longer need to suppress DPL s influence checki
341. ng and could turn it off If you were sharing this model with others in your organization that might modify it turning off Tree based dependency checking might be advisable to prevent from getting erroneous results if they changed the model in a way that violated the logic assumptions in the spreadsheet 202 Syncopation Software Chapter 6 Asymmetric Trees 6 3 ADDING A DOWNSTREAM DECISION Asymmetric trees often involve downstream decisions that represent risk mitigation options These downstream decisions are sometimes referred to as real options A real option is analogous to a financial option With a financial option you buy the right but not the obligation to purchase or sell an underlying financial asset e g a stock at some future date With a real option you invest an initial sum of money in a project or venture which gives you the ability but does not obligate you to invest more money in the project or venture at a future date after learning something about the project venture E g you could pay for some market research on a product and subsequently decide whether to launch the product or not In this instance the real option is the launch decision You get to decide whether or not launch after learning about the market research results In our model the launch costs are significant 123 million and one might ask whether there are situations in which you suspect the product will perform poorly and you would like to kno
342. node in the model In a multi attribute model following the event columns there will be a column for each attribute In a single attribute model these columns do not appear The next to last column is the probability column and the final column is the objective function column The columns display the following for each of the endpoints of the Decision Tree e Event columns the state of each decision node chance node and controlled node in the model the column labels are the event name e Attribute columns the value of each attribute at the endpoint the column labels are the attribute names e Probability column the probability of the endpoint i e the probability of the path in the Decision Tree represented by the endpoint e Objective function column the objective function value at the endpoint 308 Syncopation Software Chapter 11 Using the Endpoint Database As you examine the Endpoint Database you will notice that there are some blank cells in the event columns Blank cells may occur if there is asymmetry in the tree Blank cells in a column indicate that the state of the event heading the column is not defined i e the event does not occur on that path For example the first few thousand rows in the Endpoint Database correspond to paths where the state of the chance event Market 2 Political Stability Future is not defined because this chance event does not occur for these paths The alternatives for Pipel
343. nominal and high value and the change in color does not indicate the exact value at which the change in policy occurs If you would like more precise information on when the change in policy occurs you could run a Rainbow Diagram see Section 3 7 to narrow down the range When a tornado diagram or any other graphical output is generated and active the Chart tab group opens with a pink context shading and outline It contains two tabs Series and Format Within the Format tab you can change several display options titles legends axis colors lines shading and overall display of the chart To format the policy change colors you can chose from color palettes provided with DPL by checking the Use palette checkbox and then choosing a palette from the palette gallery in Chart Format Color You can also choose a custom color via the bucket fill icon If you have specific color preferences such as a corporate standard color scheme for presentations you can create your own custom palette in xml format and add the file to the location indicated in the Palette tooltip the tooltip will appear when you hold your mouse over one of the palette selections You can format the selected text within the chart via the View Font group 56 Syncopation Software Chapter 2 Building Decision Analysis Models gt To add labels to your chart check the Show checkbox in Chart Series Labels The labels include the High Low input values for e
344. nstance you can be comfortable with your estimation of the probability of Nominal for Cost per Unit Over a fairly wide range it does not change the optimal strategy Note however that the Rainbow Diagram in Figure 3 38 still assumes that the value for Cost per Unit is no lower than 0 8 and no higher than 0 9 The Rainbow Diagram tests the sensitivity of the optimal policy to the probability that Cost per Unit is Nominal or High given the specified Nominal and High values of the outcomes 120 Syncopation Software Chapter 3 Decision Analysis Results b pmb jb Pe e j e e a N Ke N W A A DN O WO O 10 Expected Value ee E Ee Ee 2 FEARS BS Figure 3 38 Rainbow Diagram for p The next section discusses Probabilistic Base Case Tornadoes an additional sensitivity type 121 Chapter 3 Decision Analysis Results Syncopation Software 3 9 PROBABILISTIC BASE CASE TORNADO DIAGRAM A Probabilistic Base Case Tornado Diagram is similar to a Base Case Tornado Diagram However the Probabilistic Base Case Tornado does not remove the uncertainty from the model when calculating results To calculate the base run for the diagram DPL evaluates the full model with all uncertainty as if it were running a Decision Analysis on the model Consequently the base run or the vertical line in the tornado will equal the expected value of the model as in a Risk Profile or a Policy Tree To p
345. nt NPV Reference r_by_Yr_Growth Figure 9 15 Excel Go To Dialog gt Click OK to see where this range is This two dimensional range contains values that represent growth rates for the two energy markets The Excel spreadsheet uses these if the market growth model choice is set to 3 This option allows the user to enter a separate market growth rate in each year for each of the two markets gt Switch back to DPL without changing anything in Excel gt Click Model Links Add to add a new Excel Calculation Linked value node to the model gt In the Range Names dialog click the Create arrays based on range size checkbox and select Yr_by_Yr_Growth See Figure 9 16 DPL places the node to the right in the Influence Diagram Move it if you d like 2 2 Syncopation Software Chapter 9 Multidimensional Value Nodes Range Names Multidimensional Value xls Creating Calculation Linked Values Name Location Type Court Rows Col Include_Market_2 No Assumptions C 69 Formula 1 1 Select All Include_Market_2 Yes Assumptions 0970 Formula Initial Marketing Spend Assumptions C 33 Value Mkt 2 Growth Assumptions 041 O Value OpCosts Growth_1 Assumptions 0526 Value 1 1 1 1 1 Cancel OpCosts Growth Assumptions 027 Value 1 1 1 1 1 Price_Growth_1 Assumptions C 24 Value Sort by location Price _Growth_ Assumptions C 25 Value Stark year Assumptions C 29 Value Upfront Cost Escalation
346. ntiles and Multiple Metrics Syncopation Software Display risk profile data from this model only Seres 1 aly Seres 2 iy Seres 3 2 RR e EE in dhs 4 4 4 4 Series 4 Series 5 y 4 Seres 6 Y 4 Series 7 5 8 18 g N Seres Figure 10 19 Modify Data dialog to Display Multiple Risk Profiles gt Press Ok Inthe Chart Format Legend group check the checkbox next to Show You can use the other checkboxes and drop down list within the group to format the placement and display of the legend The Risk Profile should look something like Figure 10 20 298 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics 100 90 m 80 70 60 50 CF_2014 CE 2M Aii CF 2016 Cumulative Probability 30 20 10 50 40 30 20 10 0 10 20 30 40 50 CF_2014 Figure 10 20 Risk Profiles for 2008 2010 Cash Flows You can see from the Risk Profiles that the probability distributions for 2015 and 2016 cash flow are very similar each cash flow has about a 30 chance of being zero and neither will be more than 60 The probability distribution for 2014 cash flow is quite different since this is the year when the initial investments are made As you saw earlier the net cash flow in this year will very likely be negative and could even be less than negative 50 You can close your Workspace without s
347. nty comes first in the decision tree sequence The easiest way to do this is by adding an arc gt Click Model Influence Arc Add The cursor changes to the begin arc cursor gt Click on Share of at Risk Revenue Lost The cursor changes to the end arc cursor _ gt Click on the Spin off decision DPL reorders the Default Tree to reflect the new arc Figure 5 36 179 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Shae of at Gm Trondiine Margin an pale Settlement ISERE LUY Figure 5 36 Reordered Decision Tree The Monte Carlo options in the Home Run group should be modified as shown in Figure 5 37 Make sure Risk Profile is checked Uncheck Init Dec Alts Change Initial number of samples to 500 Check the Policy Summary checkbox Select 3 from the Number of Levels drop down list 180 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models FJ Risk Profile T Policy Tree Levels 3 i Monte Carlo Samples Zal T E Objective Function W Policy Summary Endpoints Initial 500 k Decision x 7 Analysis L Init Dec Alts VOIC Value Correlations Restart 100 Run Figure 5 37 Monte Carlo Simulation Options for Downstream Decision gt Click Home Run Decision Analysis or press F10 The run may take several minutes DPL must sample each alternative of the downstream decision 100 times
348. numbers as shown in Figure 11 22 Endpoints can be exported with state names as displayed in the grid or i with state numbers for each endpoint Note when IMPORTING endpoints state numbers must be used If you intend to re impeort this file or are using it as an example for the import format export with state numbers Export endpoints with state names Figure 11 22 Saving Exported Endpoints Click Yes to use state names DPL saves the file Endpoints1 csv 331 Chapter 11 Using the Endpoint Database Syncopation Software You can open this file in Excel or in other applications and see how it has stored the endpoint data for you As shown in Figure 11 23 the rows at the top of the worksheet describe the number of endpoints and other information about the database including the state names for each event Following that is a row with column labels and then the 32 112 rows of data that you have examined previously CI I Endpointsl Microsoft Excel File Home Insert Page Layout Formulas Data Review View A o ep X B amp Calibri 11 General g Inserty E AF J By B Z U Ex 9 3 Delete g 4 Paste D A 5 2 F 3 Sot amp Find amp F Li _ Poa 00 0 Es Formate lt 27 Filter Select Clipboard G Alic e Number r Cells Editing Al fe Number of endpoints B G D Number of endpoints l 32112 Number of events 13 m gt ij lt Number of attributes 1
349. o the Unit_Sales input cell in the spreadsheet This node will represent total sales in both time periods LUU Click the Model Links Add icon Note that the default add link node type indicated by the icon is currently Excel Calculation Linked If it is not use the Model Links Add split button drop down to select Excel Calculation Linked In the Range Names dialog select Unit_Sales and click OK Double click the new Unit Sales node to edit it On the General tab change the name to Total Unit Sales On the Data tab enter the formula Sales_1 Sales_2 Use the variable button to avoid typing mistakes See Figure 3 22 Node Definition Total Unit Sales General Conditioning Print Page Setup Copy Full Screen Wheel Reorde Q Q A Data Sales_1 Sales_2 vf Links Total Unit Sales F Sales_1 Sales_2 Co ea 4 Ug Figure 3 22 Node Definition Data for Total Unit Sales Click OK to close the Node Definition dialog Move Total Unit Sales to the bottom of the Influence Diagram Draw an arc from Sales 1 to Total Unit Sales and Sales 2 to Total Unit Sales Delete the arc that goes from Sales 1 to NPV hint click near the arrowhead to select it 101 Chapter 3 Decision Analysis Results Syncopation Software Draw an arc from Total Unit Sales to NPV Your model should look similar to Figure 3 23 Marketing Price Product Price Costs 20 Strategy Change In
350. o the front as it can while keeping all other nodes fixed in position and in essence re running the model Note that DPL s solution algorithms allow it to calculate the expected value of perfect information without actually having to re run the model DPL calculates the expected value of the objective function for the modified model The expected value of perfect information for a chance node is the difference in the expected value of the objective function of the modified model and the expected value of the objective function of the base model It is called the expected value of perfect information because the modified model simulates the situation in which you have perfect information about the uncertainty before you need to make a decision For example in the base model you must make the decision to develop the product in house vs license it before you know what sales in period 1 or 2 are However in the modified model to calculate value of perfect information of Sales 1 you have the benefit of knowing what sales in the first period are before making this decision The difference in value between the modified model and the base model is due to the information gained For a given chance node if the value of perfect information is greater than zero then the optimal alternative must be different given the outcome of the chance node in the reordered tree otherwise the expected value of the objective function of the model with the reo
351. obabilities of the Operating Costs node located farther down the Decision Tree given the results of the research In this situation you can play endpoints gt Click Yes for the warning You can see from the Policy Tree or the Session Log that the expected value of the model is slightly higher than the original value 35 04 compared with 34 9 Doing research on operating costs before making the Proceed decision adds a small amount of value The added value 0 1 might be less than the cost of the research in this example This exercise gives you an upper bound on how much you would be willing to spend to gain the quality of information as specified by the Op Costs Research node Before continuing you will remove the new node Select the Op Costs Research node in the Influence Diagram and delete it Click OK for the warning You can see in the Decision Tree pane that DPL has removed the Op Costs Research node from the tree and detached the subtree beginning with the Proceed decision Select the Proceed decision in the Decision Tree and drop it on the blue triangle endpoint following the Market 2 Political Stability Now node 11 2 4 Saving Workspace Files with Endpoint Databases As mentioned earlier Endpoint Databases can require a large amount of memory and disk space when a Workspace file is saved with an Endpoint Database in it The display of an Endpoint Database also contributes a fair amount to the s
352. od variable You will need 10 additional attributes one for each year For more information on attributes please see Chapter 8 Click on Model Objective amp Utility Options i e the dialog box launcher for Model Objective amp Utility The Objective tab of the Objective amp Utility dialog appears as shown in Figure 10 3 279 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software Objective amp Utility Energy Model Objective Rename F2 Risk Tolerance Gather attribute expected values in the Policy Tree Objective function Maximize E Minimize Units for output labels Constraint function Roll forward Figure 10 3 Objective amp Utility dialog Currently there is only one attribute defined in the model It is called NPV When there is only one attribute defined in the model and no explicit objective function has been defined DPL maximizes the Get Pay expressions specified in the decision tree Therefore the objective function for the model is to maximize NPV In the model as it currently stands there are two Get Pay expressions Upfront_NPV which applies if the Proceed decision is No and Total_NPV which applies if the Proceed decision is Yes So DPL is maximizing Upfront_NPV and Total_NPV depending upon the path through the tree You will now add 10 attributes to the model for the cash flow information you are importing from Excel Click Add to creat
353. odel the situation you must tell put a Get Pay expression on the No branch of FDA Approval gt Select the No branch of FDA Approval Use the Get Pay drop down in Model Get Pay to select NPV from the list 192 Syncopation Software Chapter 6 Asymmetric Trees Objective Function ii o iv f Cost per_Unit Unit_Sales Launch Costs Marketing Costs Revenue_per Unit Figure 6 13 Model Get Pay Drop down The decision tree Figure 6 14 now indicates that DPL should add the value of NPV to the objective function if FDA Approval is No The tree has a Get Pay expression equal to NPV on the branches of the Pricing node so DPL was already adding NPV to the objective function in paths leading all the way through the tree Market Size Low Nominal Nominal High Figure 6 14 Decision Tree with Two Get Pay Expressions You can also drag and drop values from the Influence Diagram to branches in the Decision Tree to specify Get Pay expressions You will use this method later in this chapter Before you run a decision analysis we will examine an option in DPL Select Model Tree Options The Tree Options dialog will open Note that the Tree based dependency checking checkbox is checked by default See Figure 6 15 193 Chapter 6 Asymmetric Trees Syncopation Software Options Symmetric Tree based dependency checking for asymmetric trees linked to spreadsheets programs Alttemat
354. odels you will often see growth trends represented in series such as this one As this model is currently set up the series is defined in Excel rather than in DPL However in some situations you might want the series values to be defined in DPL so that they can depend on other values in your DPL model or so that you can more easily do sensitivity analysis on them You will now see how series like this one can be defined quickly and easily in a one dimensional value node in DPL Switch to DPL Select Model Links Add gt Select Mkt_1_Growth from the list See Figure 9 11 Click OK 265 Chapter 9 Multidimensional Value Node Syncopation Software Range Names Multidimensional Value xls Creating Calculation Linked Values Name Location Type Count Rowe Col lnclude Market No Assumptions C 69 Formula Select All Include Market Yes Assumptions 0970 Formula Initial Marketing Spend Assumptions C 33 Value Mkt 1_AlL Years Assumptions E 15 Formula 25 Mkt_1_Growth Assumptions C 9 Value 1 1 Cancel Mkt 2 Growth Assumptions 0410 Value 1 1 OpCosts_Growth_1 Assumptions O 26 Value 1 1 Sort by location UpCosts_Growth_2 Assumptions 0 27 Value 1 1 Bi ohew sizes Price_Growth_1 Assumptions C 24 Value 1 already linked Price _Growth_2 Assumptions 0 25 Value 1 1 Start Wear Assumptions 0929 Value 1 1 show ranges Upfront_Cost_Escalation Assumptions C 32 Value 1 1 with Formulas Yr by rr Growth Assump
355. of Spreadsheet Linking General Links Node Definition Marketing phase 2 Syncopation Software Conditioning Probability Value Print Distribution type discrete Page Setup Copy nitialization Mkt_ph2_cond_probs Full Screen Wheel Reorder M uliple Expe QAQAS ulipie Exper AAU S e000 e lt lt Marketing phase 1 Nominal High Marketing phase 2 Marketing phase 2 Marketing phase 2 Mkt_ph2_cond_probs 8000 Nominal High Low Nominal High Low Nominal High 20000 20000 8000 20000 20000 8000 20000 NANNAN gt Sale Help As noted above in a discrete chance node definition probabilities and Figure A 12 Initialization Link for Conditioned Chance Node values are initialized separately so if both will be initialized using links you need to define two range names in the spreadsheet one for probabilities and one for values In the examples above only the probabilities are linked Values can be linked in the same way simply use the Links button to fill in the Value data in the Node Definition dialog Initialization links can be very useful if you are developing large models especially if you are using your DPL model as an engine for multiple Excel templates A 4 MANAGING SPREADSHEET LINKS Once you ve established links for your individual nodes you can make changes to all
356. of a DPL program and referring to them later in the program with the verb perform A label is an identifier followed by a colon DPL determines which identifiers represent labels from their context in the sequence section All labels must be unique These labels are the equivalent to Perform Targets discussed in Chapter Section 7 1 decide to Get Info Yes and pay Info Cost then Gamble on Market Forecast then Relrdecride to Release Product Yes then gamble on Demand and get Profit to Release Product No to Get Info No then perform Rel A label must appear in a DPL program before a perform clause that references it 15 5 5 Ending a Path in a Decision Tree DPL provides two equivalent keywords for ending a path through a Decision Tree The keywords are quit and stop You can use whichever keyword makes the program easier to read In some circumstances you must use quit or stop to end a path through a Decision Tree In most circumstances these keywords are optional 416 Syncopation Software Chapter 15 DPL Programs and Code 15 5 6 Changing the Structure of a Tree with Dont Occasionally you may wish to temporarily remove a lottery from the sequence section of your model by attaching a dont gamble specification To dont a gamble simply type the word dont no apostrophe in front of the keywords gamble or take a chance on chance A 3 2 3 10 20 30 sequence dont gamble on A and get A This tree now has o
357. of attributes must produce a number between 1 and 1024 after truncation to an integer If the number is not an integer DPL will truncate it If you do not specify the number of attributes DPL assumes there is only one attribute DPL provides predefined identifiers for the attributes which you can use in the expression for the objective function In this expression 1 represents the first attribute 2 represents the second attribute and so on If you do not provide an expression for the objective function DPL uses a default function equal to the sum of the attributes If you provide an expression it need not contain an identifier for each attribute and any attribute identifier can appear more than once When the identifier for an attribute is missing from the expression that attribute does not affect the optimal decision policy The predefined identifier 0 may be used to refer to the joint probability of the path for which the function is being evaluated When you specify more than one attribute for a decision problem you must provide an expression for each attribute whenever you associate a gain or loss with an event state The expressions follow the keywords get receive pay and lose The expressions are separated by commas 419 Chapter 15 DPL Programs and Code Syncopation Software The following is an example of a DPL program that specifies two attributes value P 0 02 value MeltCost 1 0e6 decision Mode Normal Emerge
358. of the Get Pay expression you can do this via the Get Pay tab of the Branch Definition dialog Double click the branch for which you d like to copy a portion of the Get Pay select a range of cells in the Get Pay Definition dialog and press Ctrl C Then double click the 287 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software branches where you wish to paste the Get Pay expressions select the appropriate cell in the Branch Definition dialog and press Ctrl V You have now specified a value for each of the model attributes on both branches of the tree that require a Get Pay expression You are ready to run Time Series Percentiles gt If you wish save your Workspace under a new name of your choice This is not essential but you may find it useful for future reference if you intend to use Time Series Percentiles frequently gt Click Home Sensitivity Time Series The Time Series Percentiles dialog appears In the Time Series Percentile dialog you need to tell DPL which is the first and last attribute to include in the run This is done by setting the From and To drop down lists Inthe From drop down list select CF_2014 Type 2014 into the Initial time period edit box The dialog should now look like Figure 10 10 Time Senes Percentiles Energy Model Time periods attributes Percentiles From eF 20m4 o Value Type Tor cF 2023 0 i ay Port Initial time period yy g0 Paint
359. oments skewness and kurtosis and selected percentiles of the distribution For example the percentiles table shows that the 10 percentile of the distribution is 2 7 the objective function will be less than or equal to 2 7 with probability 10 Similarly with probability 90 the objective function will be less than or equal to 218 8 or conversely there is a 10 probability that the objective function exceeds 218 8 You may enter other percentiles into the Percentile list box and then click Update Percentile Values gt Click Close to close the Statistics dialog If you would like the 10 50 and 90 percentile values displayed directly on the Risk Profile select Series Labels Show 90 Syncopation Software Chapter 3 Decision Analysis Results 3 3 PoLicy SUMMARY We will now examine DPL s Policy Summary A Policy Summary is often used to understand the decision policy for a downstream decision i e a decision that occurs after one or more chance nodes so before running this output you ll add a downstream decision and related chance node to the model Right mouse click the item for the model License vs In house prob or whatever you named it earlier in the Workspace Manager and select Duplicate from the context menu Rename the new model License vs In house downstream In order to add the nodes mentioned earlier re arrange your Influence Diagram to give yourself some space between
360. ommunication of the decision structure 195 Chapter 6 Asymmetric Trees Syncopation Software 6 2 USING MULTIPLE GET PAY EXPRESSIONS Currently the model only imports data from Excel via one cell DCF Total_NPV This is the simplest way to set up a linked model and for many situations it s the best way However it is also possible to import data from several places in the spreadsheet Doing so may result in a model which is more flexible and efficient and can also help communicate the model s logic In this section you ll construct a Pay as You Go PAYG model where costs and revenues are placed in the tree to reflect when they occur Instead of using one value NPV for all cash flows associated with the project you ll separate cash flows into three components development costs launch costs and revenues marketing costs once the product is on the market You will start by creating two new value nodes for development and launch costs gt Switch back to the Asymmetric model Ctrl F12 or double click on its item in the Workspace Manager Click in the Influence Diagram pane cy Click Model Links Add Excel Calculation Linked should be the current default cy In the Range Names dialog select PV_Development and PV_Launch hold the Ctrl key when clicking on the second value See Figure 6 16 196 Range Names C Program Files x86 Syncopation DPL8 Examples Asymmetric Trees xls Syncopation Software
361. on Tree on the right Syncopation Software Chapter 1 Introduction 1 4 4 Document Navigator In addition to the Workspace Manager all the documents are also displayed as tabs in the Document Navigator see Figure 1 3 just below the Decision Tree pane You can click the tab for the item once to activate it or press the to delete the item If the desired tab is not visible in the Document Navigator use the arrows at the far right to scroll through the document tabs l gt License vs In house prob X 2 gt icense evs Figure 1 3 Document tabs in the Document Navigator Chapter 1 Introduction Syncopation Software 1 4 5 Zoom Controls You can zoom the Model Policy Tree and Policy Summary Windows by using the Zoom controls located on the status bar just below the document navigator at the bottom right of the application window see Figure 1 4 For a description of the Zoom controls see Table 1 1 j 1 oe tot a ft a TE fe 1 3G 23 a Fa i i i i i ar r Figure 1 4 Zoom Controls on the Status Bar with Zoom Dialog Active 10 Syncopation Software Chapter 1 Introduction This button launches the Zoom dialog as seen in Figure 1 4 within which you can select a preset level of zoom or specify a custom zoom level Zoom the active pane so that all of the Influence Zoom Full Ne i Diagram Decision Tree is visible in the pane Zoom Percentage ZOOM Click the Zoom out
362. one policy change for each bar and that Sales 1 and Sales 2 have the biggest impact on the objective function Notice that the vertical line is at 111 1 which as noted in Section 3 3 is the expected value of the objective function of the model as currently built You have now reached the end of the tutorial on building and analyzing DPL decision analysis models Though the model you developed in this tutorial is fairly simple you now have the essential skills needed to develop models of decisions varying complexity 123 Chapter 3 Decision Analysis Results Syncopation Software 124 Syncopation Software Chapter 4 Strategy Tables 4 STRATEGY TABLES When building a decision model you may find that there are a number of up front initial decisions that should be included in the model but that not all of the combinations of the decision alternatives make sense or are viable Initially it may be easier to think about the decisions individually when framing the problem and developing the model However when it comes to analyzing the model using a decision tree with a large number of separate decisions may not be either practical or desirable In situations such as this you can use a strategy table node to aggregate your up front decisions into a manageable number of strategies In this context a strategy means a specific combination of decision alternatives for the up front decisions in the decision model This chapter assumes you
363. ons of how DPL performs various calculations and how to interpret each of DPL s outputs are also in online help Online help also contains a listing of common error and warning messages that may be generated by DPL You can access online help by using the commands within the Help group of the Help tab on the ribbon or by pressing F1 The latter provides context sensitive help which means that pressing F1 will open online Help to the topic corresponding to what you re doing in DPL You can also obtain technical and installation support by emailing support syncopation com Support resources such as technical notes case studies and answers to frequently asked questions are available on our website www syncopation com Perpetual licenses of DPL include 90 days of Premium Care to help you get started with DPL Premium Care includes phone and email support feature releases and upgrades to major product releases at no additional charge and allows you the ability to create support tickets on our website Following the first 90 days Premium Care can be purchased on an annual basis We recommend you consider purchasing Premium Care More information is available at www syncopation com or by emailing Sales syncopation com Annual licenses include Premium Care The quickest most efficient way to become proficient with DPL is to attend a training course Syncopation offers DPL training courses a few times a year at various locations Custom training
364. onsists of two Base Case tornadoes on one chart The diagram shows that the chance event Unit Sales has the greatest impact on the objective function value regardless of which initial alternative is pursued It also shows that the impact on the objective function of Unit Sales is less for the License alternative than it is for the In house alternative the bar for License is narrower Also the diagram shows that if the License alternative is pursued despite the fact that it is not the optimal policy the chance events Marketing Costs and Cost per Unit have no effect on the objective function This makes sense based on how the model is set up If the product is licensed there are no marketing costs or per unit costs in the calculation of NPV in the spreadsheet 71 Chapter 2 Building Decision Analysis Models Syncopation Software Initial Decision Alternatives tornadoes are helpful for comparing the impact of each chance event on the objective function across the possible alternatives of the initial decision in the model Although the tornado you ran in this section is simple and illustrative Initial Decision Alternatives tornadoes can be quite powerful when used with models that incorporate several strategic alternatives and several chance events Note in the simple example model there is only one decision in the model Since DPL runs an Initial Decision Alternatives Tornado by fixing the initial decision at each alternative and running a Base
365. ook like Figure 5 30 173 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Spin Off Trendline RMC Shae of at knes Revenue Lawsuit can Growth Rate Settlement CE Figure 5 30 Model with Spin off decision m Note Loss of Synergy is simply a linked value node that is set to 0 06 for now You are linking this value to the model because in the next section you will see how to perform sensitivity analysis on the assumed value You are ready to run the decision analysis model You may want to save your workspace at this point Change the initial number of samples to 1000 gt Make sure Risk Profile and Init Dec Alts is checked The Home Run group should look like Figure 5 31 Risk Profile T Policy Tree Levels none Monte Carlo Samples i Objective Function 0 Policy Summary Endpoints Initial 1000 k Decision Analysis M Init Dec Alts O voic C Value Correlations Restart 100 X Rur Figure 5 31 Home Run options for Monte Carlo Simulation Click Home Run Decision Analysis or press F10 174 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models The analysis will take a few seconds Excel has to recalculate the spreadsheet 2000 times 1000 for each decision alternative When the run is finished DPL will show you the outcome distributions for both alternatives gt Uncheck Histogram in
366. opeten Sotware Pe L w Low Low Low Low Low gt cash flow decidon X 4p Fee Help press Fl oome 9 Full Prev 4 Figure 16 1 Software da Workspace Inthe Home Run group in addition to the outputs already selected check the checkbox next to Endpoints and select Full Tree Enumeration for the evaluation method Run the model to see how long it takes to run linked to the Excel spreadsheet gt After the model runs look at the Session Log to see how long the model took to run Near the bottom of the Session Log will be a line with Elapsed time Note down the number Double click the item for the model in the Workspace Manager to activate the Model Window The model is linked to an Excel spreadsheet called Cash Flow1 xls To convert a linked model you use the Model Links dialog Select Model Links Options The Model Links Dialog appears with the Link tab active as shown in Figure 16 2 426 Syncopation Software Chapter 16 Converting Spreadsheets to DPL Code Model Links cash flow decision links Extemal Links Settings Spreadsheet Database Name Cash How xls Calc Init Nodes 17 0 of 17 Browse Remove Show Nodes Convert to Cale Program Create Link to New Spreadsheet Links to Programs in the Workspace DPL Program for Data Definitions DPL Calculation Program Values defined in this program can be used in both the influence diagram and the decision
367. ord cccceeeeeeseeeeeeeeseeeees 417 Get Pay expressions adding tO tree csceeeeeeeeeeeeees 192 COPY and PASTE ceceeceeeeveeees 287 deleting irasi nannaa 209 using MUItIPle cceeeeeeeeeees 196 given KEYWOTFK cecceceeeeeeeeeeeeeaees 402 if then else directive n 423 imperfect information 216 include directive ccceseeeeeeeeeeees 422 influence arcsS sssssssssnessssnnesseeresne 32 arrowhead COIOMS sccecsseeeeees 232 DEMGING conaran 46 changing tyPES cccecceeeeeeeeees 223 create short CUt ccccceeeeeeeeeeees 40 default types ccsccecseeeeeeeeees 222 SUMMA Y onei a 232 UNO eaaa 103 Influence Diagram pane 8 Initial Decision Alternatives tornado 70 initialization links EStADIISNING ccceceeeseeeeeaeeees 445 for conditioned nodes 449 Integer keyword ccsceeeseeeeeeeeees 411 keyboard shortcuts cecsecseeeeees 457 SE E E T es 215 Learning CVeNtS ccccceeeeeeeeeeeeaes 319 mixed alternative outcome grouping E E E E T EE 210 Model WiNGOW sscsccseseeseceeseeeeeeees 8 SPIRET arinin 8 models DUILAIG nanira 23 Monte Carlo simulation 00 141 MUA arrinin See multiple attributes multidimensional value nodes 253 In Get Pay expressions 300 multiple attributes ccccceeeeeees 235 adding attributes cccceceeees 241 in Get Pay expressions 4 243 Policy Tree WiIth
368. orkspace Manager window right click on Model1 and select Duplicate Figure 5 15 Workspace Manager 7 EG Workspace Tor Deck Properties Delete Rename K Duplicate Add to Workspace Portfolio Update Report Add Chart Figure 5 15 Using the Workspace Manager to Duplicate a Model DPL makes an identical copy of Modeli Rename the first model Deterministic by right clicking on it and selecting Rename Rename the second model Risk Analysis in the same way cy Activate the Risk Analysis model If it isn t already double click on the splitter bar to maximize the Influence Diagram pane cy Click Zoom Full see Table 1 1 button on the status bar or press Ctri L Your Workspace should look like Figure 5 16 157 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Gud DPL Enlightity Risk Anatysis ad S Nome Mode View Dete Meto e instance bocctive F tion O S amp x Mar 7 i 4 Straighten gt ontrol alie Uncortros All Eat eet may exeeession Add Edit Change Ade Add From Sestion log 9 OPL Decaion Programming Lanqueje a Copyright 2033 2013 by Syncopeten Sofware Mhceroer ating sofware an icense 5 Syncopaton Sofware inc Al rgs reserved Cuiside he US 1 070 230 2568 Professene Versen Remane E 00 DO Senal Nurser SYOPOO120102001 Licenses User suser Syncopation Software inc 13 4616 Comping Model 13 40 16 Cor
369. ormally would run a model Home Run Decision Analysis or press F10 The program will open the linked Excel spreadsheet and run the model producing the same initial results you saw in Chapter 11 The expected value of the model is still 34 9 You will make a change to the program that affects the value model and run it one more time gt Click in the Program Window to edit the program Inthe second line of the program change the discount rate from 0 10 to 0 05 The code should now look like this excel Excel 1 ASsumptions Discount Rave value Discount Rate 05 gt Run the program again and inspect the new results The results are quite different because a lower discount rate implies a higher value for investments with long term payback The expected value is now roughly 57 and the optimal policy is always to proceed with the pipeline strategy regardless of the outcomes of the initial uncertainties Keep in mind that the program and graphic model are not linked so changes in one do not appear in the other Also it is not possible to go the reverse direction and generate a graphic model from a program Also note that the program in this example is still linked to the Excel spreadsheet for evaluation of the cash flow results That is when you ran the program the program had to open the Excel spreadsheet to calculate Upfront_NPV and Total_NPV just as the DPL model would do 398 Syncopation Software Chapter
370. orporated into it There are now two models in the Workspace DPL will always run an analysis on the model you are viewing or on the model whose output you are viewing If that model is not the bold model in the Workspace Manager DPL will make it the bold model 66 Syncopation Software Chapter 2 Building Decision Analysis Models 2 7 7 BASE CASE TORNADO DIAGRAMS You are now ready to run another sensitivity analysis on your model the Base Case Tornado Diagram The Base Case Tornado diagram is designed for models which have several discrete chance nodes each with a plausible default state the Nominal or Base Case value and most of which have three or more outcomes To compute a Base Case Tornado Diagram DPL first sets all discrete chance nodes to a state that you define to be Nominal for each node to calculate a Base Case It then varies each discrete chance node individually to a state you specify to be Low and to a state you specify to be High while keeping all other discrete chance nodes at their Nominal states The Base Case Tornado Diagram is useful for understanding the impact of each discrete chance node on the model s objective function To run a Base Case Tornado Diagram drop down the Home Sensitivity Tornado split button and choose Base Case from the list The Base Case Tornado Setup dialog appears as shown in Figure 2 37 Base Case Tornado Setup License vs In house prob Bars Chance Low State Nominal State
371. ow use DPL s Filter by Decision Alternative feature to give you information on which event states are correlated with drilling Activate the Policy Tree again Select Policy Policy Summary Compare By Decision Alternative The Filter by Decision Alternative dialog appears as shown in Figure 12 13 Filter by Decision Alternative The baseline for comparison is the set of endpoints that are included in the optimal decision policy Filter the endpoints by selecting those endpoints that are included in the optimal decision policy AND for which the decision below takes the selected alternative Decision Drill Alternative Ves Figure 12 13 Filter by Decision Alternative Dialog The dialog allows you to select which decision and alternative you would like to use for the filter As the dialog indicates the endpoints will be filtered to select the endpoints that are part of the optimal set of endpoints AND for which the selected decision takes the selected alternative gt The decision and alternative in which you are interested Drill and Yes are already selected Click OK DPL creates a new filtered Policy Summary as shown in Figure 12 14 Note that the original Policy Summary is still in the Workspace Manager it was not overwritten 355 Chapter 12 Advanced Decision Analysis Results Syncopation Software Test Drill Amount of Oil i None lt 0 gt D lt 32 gt 0 Yes lt 100 gt 16
372. pears if advanced features are turned on define risk tolerance by branch Display tab change get pay and state display options The Get Pay tab of the Branch Definition dialog will open if you double click the branches of a node in the decision tree or click Model Get Pay Options Table 1 5 Model Branch Ribbon Group 16 Syncopation Software Chapter 1 Introduction Select Attributes Drop down List Choose the attribute from the drop down list for which you d like to edit the get pay expression for the selected branch es Only necessary in multiple attribute models Objective Function x Edit Get Pay Combo Box o vf Use the drop down list to select a single quantity or type in an Show _ Hide Indicator l i i 7 expression for the get pay for ened the selected branch es For multiple attribute models the Get Pay tab of the Branch Definition get pay will be for the attribute The dialog box launcher will open the selected in the Attributes drop dialog See dialog box launcher a down list otherwise it is for the description in Control column for the Lz Branch group in Table 1 5 objective function Use the h or buttons to insert a variable or function Get Pay Display Radio Buttons Display hide or show an indicator for the get pay for selected branch es Table 1 6 Model Get Pay Ribbon Group 17 Chapter 1 Introduction Syncopation Software Objective Function Com
373. port node which is typically linked to a cell with a formula in it You can check or uncheck these options to see the full set or restricted set of cell names if you need to 444 Syncopation Software Appendix A Overview of Spreadsheet Linking A 2 4 Summary of Calculation Links Methods Table A 1 summarizes when each method is appropriate Building a spreadsheet linked Method 1 DPL model from scratch Adding a node to the DPL model Method 2 for a new driver cell in the spreadsheet Linking an existing or newly Method 3 created node to a cell in the spreadsheet Table A 1 Summary of Methods for Establishing Calculation Links A 3 INITIALIZATION LINKS Initialization links are appropriate when you wish to store probabilities and values for DPL node data in a spreadsheet DPL uses the initialization links to set the probabilities and values for the linked node s Initialization links are different from calculation links in that DPL gets the data from the spreadsheet once at the beginning of a run to initialize the node s values or probabilities whereas calculation links exchange data to from Excel during the entire course of the run To use initialization links first define a range in Excel that contains the probability or value data you wish to link to the node in DPL The way you define the range will vary depending on the node and its conditioning For a simple 3 state chance node you define a 3 cell range containing the p
374. puts Use the Risk Profile drop down list just below the Risk Profile checkbox to select Objective Function and All Attributes Obj Fn amp All Atts This will generate the full Risk Profile for the objective function and every attribute in your model See Figure 10 16 gt Click Home Run Decision Analysis or press F10 gt Click Yes to the warning a W Risk Profile W Policy Tree Levels 14 all at Obj Fn amp AN Atts gt Policy Summary Endpoints Decision Analysis Y Init Dec Alts U YOI Value Correlations Run Figure 10 16 Generating Risk Profiles for All Attributes The model may take a few minutes to run As mentioned earlier adding attributes to a model does have a performance impact however requesting a single Risk Profile vs multiple Risk Profiles does not When it completes you will see Risk Profile charts and datasets for the objective function plus all the attributes in the Workspace Manager as in Figure 10 17 295 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software E UEI DE DPL Tene Senses Pescenbles da Policy Tree a Rasa Toh Pay Tit Wie d ay a Aga Cangi JP Gepected Value Jy Pity Tiit Session Log DPL Deets Programing Langoase Saar 0 by Syncoeaied Ge Merar mooning ofre Oo Boe io capaho Safwan he fl ogtie reer BC FET CEES aad uppi rhon e ui itfa 1 pbi ai 27 iniside the US t 97A 210 2509
375. r are conditioned by other nodes You will start by creating an influence arc between Revenue per Unit and NPV To create an influence arc click the Model Influence Arc Add button on the ribbon see Table 1 3 The mouse cursor changes to the begin arc cursor Place the cursor over the Revenue per Unit node and click Now the cursor changes to the end arc cursor and an arc is anchored at the Revenue per Unit node Revenue per Unit is called the predecessor node for this arc Place the cursor over the NPV node and click An influence arc is drawn from Revenue per Unit to NPV See Figure 2 9 NPV is called the successor node for this arc 32 Syncopation Software Chapter 2 Building Decision Analysis Models Figure 2 9 Model with Influence Arc An influence arc that points from node A to node B indicates that node B depends on is conditioned by and or occurs after node A In Figure 2 9 the influence arc you just added indicates that NPV depends on Revenue per Unit Note DPL indicates which arc is selected by coloring it magenta In general DPL indicates which item is selected in the Influence Diagram or Decision Tree by coloring the item magenta There is an easy way to have DPL automatically generate influence arcs based on your formulas or spreadsheet links gt Click Model Influence Arc From Formulas on the ribbon see Table 1 3 DPL generates the influence arcs from the formulas found in the
376. r each conditioning event For example if value node C were conditioned by discrete chance node A and decision B in the Conditioning section of the Select Value dialog you would need to specify an outcome for A and an alternative for B in order to run a Two Way Rainbow Diagram on C Select FIRST Value x axis for Two Way Rainbow Value for sensitivity Per Unit Costs Conditioning 1 Per_Unit_Costs Nominal Current value 0 8 Cancel Figure 14 4 Select Value Dialog Displaying Conditioning for Per_Unit_Costs In the drop down list in the Conditioning section select High Click OK The Run Two Way Rainbow Diagram dialog updates to indicate the value and the conditioning states if any that you have selected DPL displays a value and its conditioning using the notation value event state You have selected the Per_Unit_Costs value for the High state of the Per_Unit_Costs chance event so DPL displays Per_Unit_Costs Per_Unit_Costs High in the Variable name label box If the value you have chosen is conditioned by multiple events this is indicated by separating each event state with commas i e value eventi state event2 state etc In the Current value label box DPL displays the current value for the variable given its conditioning See Figure 14 5 378 Syncopation Software Chapter 14 Sensitivity Analyses Two Way Rainbow Diagram License vs In house downstream First Variable 4 axig Variable na
377. rdered tree would be the Same See the Policy Tree from the reordered model in Figure 3 28 In the example model the value of perfect information for Sales 1 is approximately 3 therefore the optimal alternative must have changed for one of the outcomes of Sales 1 If you had the luxury of knowing which state Sales 1 was in before deciding on Product Strategy you would choose to License in certain scenarios Overall In house is the better alternative but with information on Sales 1 you would choose to License for those scenarios in which License is optimal 108 Syncopation Software Chapter 3 Decision Analysis Results Price Change In house 40 7 Product Strategy Cost per Unit Low 51 1 Raise 42 8 30 l Price Change License 51 1 Cost per Unit Keep Same 51 1 Cost per Unit Raise 106 6 Price Change In house 107 7 Sales 1 Product Strategy Cost per Unit 114 2 Nominal 107 7 Keep Same 107 7 40 Price Change License 93 7 Cost per Unit Raise 186 0 Price Change In house 186 0 Product Strategy Cost per Unit High 186 0 Keep Same 179 3 30 Price Change License 139 2 Figure 3 28 Policy Tree to Illustrate Value of Perfect Information DPL calculates the expected value of control for a chance node in a similar manner to the way in which it calculates the value of perfect information except that in addition to moving the chance node to the front of the tree it also turns it into a dec
378. rdinates represent outcome values and whose Y coordinates represent the sum of the probabilities of all possible outcomes less than or equal to the associated outcome value In the Home Run group you can choose to have DPL generate and display cumulative distributions for the optimal policy or all initial decision alternatives Also if you have defined more than one attribute for the model a separate cumulative distribution for each of the attributes can be generated The cumulative distributions are displayed in a Risk Profile Chart Current Model The model that has been most recently run or compiled The current model s item in the Workspace Manager is indicated by bold face type Cycle A set of nodes and influence arcs that create a loop in which a node depends directly or indirectly on itself An Influence Diagram may not include cycles 462 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms D DDE See Dynamic Data Exchange Decision Alternative See Alternative Decision Node A node in the Influence Diagram or decision tree representing an event that the decision maker has control over i e an event with alternative choices for the decision maker to choose among During evaluation the alternative that maximizes or minimizes the objective function is chosen The graphical symbols for a decision node are a yellow rectangle in the Influence Diagram and a yellow square in the decision tree De
379. re is an equal chance that the time series variable is less than or greater than that amount The expected value series is the expected value of the variable for the time period Finally you can change several display options to tailor the appearance of the Time Series Percentiles chart to suit your needs With the Time Series Percentiles chart still active uncheck Error Bars in Chart Format Display In Chart Series Lines Markers make sure Lines and Dashed are checked The chart will look like Figure 10 13 291 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics 30 25 o qe an T Se a emer a 20 i r 15 j 10 j ui re thy aim Ss Pes az i R rr no O tem aN 3 i if T O a E E Ga Gee 0 E lt gt 0 i i i 5 a EE 90th Percentile 10 mf f 50th Percentile E 10th Percentile 15 T S Expected Value Li t 20 H i a H 25 A j 30 H 1 i 35 H i I 40 i i 45 50 T T T T T T T T T T 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 Figure 10 13 Revised Time Series Percentiles Chart The revised chart connects the points for each series 10 percentile etc over time Technically how the uncertainty in the attributes evolves between time periods cannot be derived from this output That is the percentil
380. re Unstable excel Excel _ 1 Assumptions Operating Costs chance Operating Costs Low Nominal High 3 4 3 18 Operating Costs Low 20 Operating Costs Nominal 22 Operating Costs High chance Appraisal Results Low Nominal High Initial Upstream Investment Actual Reservyes Initial Upstream Investment Low eons ee ok Actual Reserves Low Veo ge Aon 2 Oe Actual _Reserves Nominal 25 3 45 Actual Reserves High Initial Upstream Investment High 2 Oy ol pr Oo hy Actual _Reserves Low ie Oe E yams ON Som ar Actual _Reserves Nominal 2057 sie ode Actual Reserves High excel Excel L Assumptions Total NPV value Total NPV excel Excel 1 Assumptions Upfront NPV value Upfront NPV Sequence decide to Initial Upstream Investment then gamble on Appraisal Results then Gamble on Forecast Price then gamble on Market 2 Political Stability Now then decide Lo Proceed res then decide to Pipeline Strategy Mkt 1 10 BCM and set Pipeline Capacity BCM 10 and set include Market 2 No and StrategyTableSubtree0 av gamble On Act iak Price then gamble on Actual Reserves then gamble on Operating Costs then gamble on CapEx and get Total NPV to Pipeline Sirategy Mke 1 20 BCM and set Pipeline Capacity BCM 20 and sex 0C Lude Markot 2 No and perform StrategyTableSubtree0 tO Pipeline Strategy Mkt L and z 10 BCM and set Pipeline Capacity BCM 10 and set Include Market 2 Yes and Strat
381. refers value B3 1 2 to Value Bi le Value B2 Zs value B3 Bl B2 Uncheck this option if you want the constant values As above but pertaining to formulas Check this to have DPL ignore i e not convert cells or ranges which are named but not used in other calculations This saves some memory but can be annoying as output cells like Total_NPV are often unreferenced DPL series are designed for fast calculation of similar formulas as are often found in spreadsheet rows However a whole series row must be evaluated at once so DPL series can be recursive even though the source spreadsheet has no circular references Check this option if you see the message series definitions involving X are recursive DPL will generate less efficient code using values and arrays Convert e g NPV on the DCF sheet to DCF_NPV rather than just NPV This should be on when using Convert to Calc Program from the Model Links dialog Table 16 2 Code Generation Conversion Options definitions 430 Syncopation Software Syncopation Software Chapter 16 Converting Spreadsheets to DPL Code On Suppress warnings Uncheck to see a message box for every warning Replace with Unsupported functions are na rather than normally replaced with na This 0 is a safety feature if the expression needs to be calculated an error will stop the run Uncheck this to have unsupported functions replaced with zero Ta
382. rfect Information Control VOIC Endpoints and Value Correlations are not checked Click the Home Run Decision Analysis icon or press F10 DPL runs the analysis and produces the requested outputs 77 Chapter 3 Decision Analysis Results Syncopation Software 3 1 Povicy TREE The Policy Tree output will be active after the run is complete as shown in Figure 3 2 Before discussing the Policy Tree take another look at the Workspace Manager Note that DPL has put double chevrons around the License vs In house prob model you just ran This indicates the model is compiled i e checked for errors and data structures loaded in memory and is ready to run Once you modify the model the compiled indicator is removed ous hla TOA fees Pelee alegre PR Toa 5EN EA 4 Unit Sales In house 107 7 Product Strategy 107 7 B EEEE A Unit Sales Cepia E SS ee Segregated Se Pen penaing bee oi eee y License 93 7 O Lenin vi rbaa Op Lene va inher prob X Temata Tmt Tamea Pokey Iie F Titi af ii Deien Alteration x Figure 3 2 Policy Tree for Probabilistic Model The Policy Tree indicates that In house is still the optimal decision alternative with a thick branch The expected value of the model has changed from 105 1 to 107 7 Note that this is the first time that you have a true expected value that incorporates probabilities since there are now chance events in the
383. rgiete 13 54 00 Runsing Vave Tornado for Bad Dett As a Deterministic K Teensdol X gt Rink Analys X Far Help press FI ooms 147 Fal Pres l Figure 5 16 Workspace with Two Models You may notice that the tornado diagram is under Deterministic in the hierarchy DPL groups outputs with the models they came from It s good practice to save your Workspace from time to time Select File Save Workspace 158 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models 5 6 RUNNING A MONTE GARLO SIMULATION The tornado diagram indicates which values have the highest impact on Total Value The next step in our risk analysis is to model the uncertainty in those factors You ll start with Gross Margin since it s at the top of the tornado Right click on Gross Margin select Change Node To and then Continuous Chance in the context menu The Node Definition dialog appears as shown in Figure 5 17 Node Definition Gross Margin Name and settings Data Gross Hide Margin Links DPL variable name for expressions Gross_Margin Comments Print All Comments Copy to Clipboard cant He Figure 5 17 Node Definition Dialog Gross Margin is now a chance node meaning that it is uncertain DPL has two types of chance nodes discrete and continuous Discrete nodes are used for uncertainties which have a finite number of outcome
384. rious combinations of states is sometimes called row major order or the odometer principle This means that the states of the last event vary most rapidly The same applies for the values 402 Syncopation Software Chapter 15 DPL Programs and Code The comments in the earlier statement show the states of the conditioning events These comments are ignored by DPL and they do not determine the order in which you specify the probabilities 15 4 6 Controlled Events The following statement defines a controlled event in DPL controlled AchieveOrbit Yes No If you wish to associate value expressions with the states of a controlled event enter them after the state names controlled AchieveOrbit Yes No 500 10 15 4 7 Values The following statement defines a value in DPL value Base Price 585 30 The following statements define two events and a value Inflation_Factor that depends on the two events chance MoneySupply Lower Constant Higher f 3p OTe chance TaxRate Unchanged Increased 1 value Inflation Factor given MoneySupply TaxRate TaxRate Unchanged TaxRate Increased 1 14 1 12 MoneySupply Lower dees Bale 1 09 MoneySupply Constant 1 09 1 08 MoneySupply Higher The order in which you list the numbers depends on the order in which you list the conditioning events DPL uses the first number when all conditioning events are in their first state The next number applies when the last con
385. rmula DPL creates import series nodes with no data For multiple cell two dimensional arrays that do not contain any formulae DPL creates export two dimensional array nodes initialized with data from the spreadsheet For multiple cell two dimensional arrays that contain a formula DPL creates import two dimensional array nodes with no data The Create Linked Values from Excel method can be used in conjunction with the Create Model from Excel method You might start by using Create Model from Excel and subsequently decide that existing named cells within Excel need to be linked to DPL or you may modify the Excel spreadsheet and then need to link these new named ranges to DPL 44 Appendix A Overview of Spreadsheet Linking Syncopation Software A 2 3 Method 3 Establishing Calculation Links for an Existing Node You may already have a node created in your DPL model that needs to be linked to your spreadsheet In this situation edit the node as usual and switch to the Links tab of the Node Definition dialog The Links tab of the Node Definition dialog allows you to link an existing Influence Diagram node to an Excel cell From the Links tab Specify Excel as the Calculation Link type DPL automatically sets the spreadsheet to the current linked spreadsheet If there isn t one use the Browse button to select the spreadsheet gt Click Cell Names Select the desired cell from the list gt Click OK Alternatively
386. rnado diagram is an extension of the Base Case tornado that you ran in Section 2 7 To create an Initial Decision Alternatives Tornado DPL sets the initial decision to each of its possible alternatives and runs a Base Case tornado for each alternative The resulting Base Case tornadoes are then shown on one diagram You will now run an Initial Decision Alternatives tornado using the model you built in the previous sections Make sure the License vs In house prob model is still the active model gt Drop down the Home Sensitivity Tornado split button and select Initial Decisions Alternative from the list The Initial Decision Alternatives Tornado Setup table appears It should be identical to the Base Case Tornado Setup table you examined in Section 2 7 Leave the settings as they are gt Click OK An Initial Decision Alternatives tornado appears as shown in Figure 2 39 70 Syncopation Software Chapter 2 Building Decision Analysis Models Initial Decision Alternatives Tornado Unit Sales A TTT ETE TTT Price Marketing Costs Cost per Unit e EG Product Strategy In house I Product Strategy License Figure 2 39 Initial Decision Alternatives Tornado Diagram The tornado shown in Figure 2 39 is an example of a very simple Initial Decision Alternatives tornado There are two initial decision alternatives In house and License so the tornado diagram essentially c
387. rned on These settings override the settings in the Model Options DPL will use the settings in the Options menu to set options that are not included in the option list The letters may be in either upper or lower case 423 Chapter 15 DPL Programs and Code Syncopation Software 15 6 5 Command Window The Command Window allows you to view any number or calculation within any scenario in a model You can view any number in a model including values probabilities series and array elements and expressions You can also temporarily change any value in a model These two capabilities provide a powerful and flexible environment for testing and debugging TO open a Command Window select Home Workspace Add to WS Command Program You will see a blank window in which you can enter various types of commands to execute A Command tab appears and becomes active on the ribbon as well The results of your commands are displayed in the Session Log Refer to the DPL online Help for detailed descriptions of specific commands 424 Syncopation Software Chapter 16 Converting Spreadsheets to DPL Code 16 CONVERTING SPREADSHEETS TO DPL CODE This chapter is intended mainly for experienced DPL users who have built large Excel linked DPL models and who need to reduce model runtime This chapter discusses how to convert a spreadsheet to DPL code provides guidance on good spreadsheet practices if you intend to convert to code and alerts you to
388. robabilities and or another 3 cell range containing the values Probabilities and values are initialized separately in chance nodes and must be stored in separate named ranges in Excel For a chance node that is conditioned by another event you define a two dimensional cell range e g 3 by 3 in which the rows correspond to the states of the conditioning node and the columns correspond to the probabilities or values for the states of the conditioned node An example of this is given below for a conditioned 3 state chance node There are two methods to create initialization linked nodes 445 Appendix A Overview of Spreadsheet Linking Syncopation Software 1 If you have not created the node yet use Model Links Add Excel Initialization Linked See Figure A 6 Range Names C Program Files x86 Syncopation DPL3 Examples Cash Flow with Decision xls a Creating Initializaton Linked Values OF Name Location Type Count Rows Col Mkt_phl_ probs Initialization C E SY Value 1 Select All MRE ph1_ values lnitializationI C 3 E 5 Value Mktphe cond probs Initialization C 17 Value Mkt phe_probs Initialization C 10 Value Mkt_ph2_values Initialization C 11 Value Cancel TOTAL_COST_OF_ SALES CFIfD0 1440 14 Formula TOTAL_OVERHEADS CF D0 20 0 20 Formula L Sort by location i l Wages salaries CFIf0 1 0 1 Formula Booe already linked Show ranges with formulas Create arrays based on range ize
389. roduce the results for each chance node in the model DPL runs the model with the chance node it is evaluating fixed to a low state while allowing all the remaining chance nodes to vary across their outcomes DPL then repeats this with the chance node fixed at a high state and repeats the whole process for each chance node Because the base run in the Probabilistic Base Case Tornado matches the expected value of the model in a Decision Analysis run it can be more easily compared to other outputs from a Decision Analysis run and sometimes easier to explain The process for setting up the Probabilistic Base Case Tornado is identical to that of the Base Case gt Drop down the Home Sensitivity Tornado split button and choose Probabilistic Base Case from the list The Probabilistic Base Case Tornado Setup dialog appears The set up is identical to that of the Base Case Setup Dialog in Figure 2 37 gt Click OK to run the Probabilistic Base Case Tornado Diagram DPL produces the Probabilistic Base Case Tornado Diagram as shown in Figure 3 39 122 Syncopation Software Chapter 3 Decision Analysis Results Probabilistic Base Case Tornado Sales 1 Sales 2 Marketing Costs Cost per Unit 60 80 100 120 140 EE Low Policy Change E Policy same as Nominal E High Policy Change Figure 3 39 Probabilistic Base Case Tornado Diagram The tornado indicates that all variables are decision sensitive there is at least
390. roduces the Probabilistic Event Tornado shown in Figure 14 11 Probabilistic Event Tornado Sales Sales 2 Per Unit Costs Marketing Costs ORIGINAL MODEL RANGE 5 10 15 20 25 30 35 40 45 50 55 60 65 Expected Value E 10th to 50th Percentile BJ 50th to 90th Percentile Figure 14 11 Probabilistic Event Tornado 387 Chapter 14 Sensitivity Analyses Syncopation Software The width of each bar indicates the difference between the 10th percentile and the 90th percentile of the risk profile of the objective function for the model when the sensitivity event labeling the bar is the only uncertainty removed from the model unless other chance events have been set to Don t Gamble The change in color occurs at the 50th percentile The bar is yellow between the 10th and the 50th percentiles red between the 50th and 90th percentiles The black vertical line running through all the bars indicates the expected value established in the run for the Original Model Range As in a Deterministic Event Tornado a bar may not change color if the risk profile is highly skewed for the model with the sensitivity event replaced by its expected value The narrower a bar is for a chance event in a Probabilistic Event Tornado the greater the reduction in uncertainty is due to the removal of the chance event Bars nearer the top of a Probabilistic Event Tornado contribute more to the overall uncertainty in a model 14 5 W
391. rs and Annual Growth Assumptions sections near the top See Figure 9 10 x iW Is Multidimensional Value xls Compatibility Mode Microsoft Excel 0 Home Insert Page Layout Formulas Data Review View DPL A Q o ee 8s B amp Arial 10 General E Conditional Formatting 3 Inserty E Ay A B3 B Z U Aa Z 7 E Formatas Table 3 Delete 7 Z Paste t A a o0 pee Sot amp Find amp FY E 227 Ar HE DP es E Cell Styles 7 Hel Format C2 Filter Select Clipboard Font Alignment Number Styles Cells Editing Al v f z E ET F G H t J K LI M N amp Basic Parameters Color coded blue yellow or green gt currently linked to DPL Discount Rate p 1 Market 1 Growth Model Choice a ieee Market 2 Growth Model Choice E Market 1 Size Year 1 o o Market 2 Size Year 1 p80 Annual Growth Market 1 Option 2 50 Annual Growth Market 2 Option 2 50 Annual Growth Assumptions for Market Growth Models 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 Market 1 Size Option 2 110 116 121 127 134 140 147 155 163 171 Market 2 Size Option 2 180 189 198 208 219 230 241 253 266 279 Year by Year Growth for Market 1 Option 3 Year by Year Growth for Market 2 Option 3 0 04 005 005 005 004 004 0 03 0 03 0 03 0 03 0 04 0 05 005 005 004 004 003 003 003 0 03 4 gt gt Assumptions DCF lt Market Trend Data lt 3 li 4 Il b i Ready amp 100 C
392. s Select the Data tab Enter 1 1 for the Raise alternative and 1 0 for Keep same These numbers will be used as multipliers so 1 1 means raising prices 10 Press Enter The Data tab should look like Figure 3 13 Node Definition Price Change General Conditioning Print Page Setup Copy Full Screen heel Reorder Q 3 Lo vis Links Price Change Keep same Figure 3 13 Node Definition Data Tab for Price Change Decision Click OK Double click Unit Sales to edit its definition On the General tab change the name to Sales 1 gt Change the value for each outcome to half its previous value The Data tab should look like Figure 3 14 when you are done 93 Chapter 3 Decision Analysis Results Syncopation Software General Conditioning Print Page Setup Copy Full Screen Wheel Reorder odaia Distribution type discrete y Probability 3 Value 10 Figure 3 14 Node Definition Data Tab for Sales 1 gt Switch to the Links tab and under the Calculation links section at the top select the None local radio button to break this node s link with the spreadsheet more on this later General Data Calculation links These links are recalculated throughout the run They may be either inputs used in the linked spreadsheet s calculations e g unit sales or calculated resul
393. s such as single events e g regulatory approval or fixed scenarios Low Base Case High Continuous nodes can have infinitely many outcomes sampled from named probability distributions normal triangular poisson etc In this chapter you ll be using continuous chance nodes DPL draws discrete chance nodes in bright green and continuous chance nodes in dark green 159 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Next you ll need to edit the data for Gross Margin gt Click on the Data tab of the Node Definition Dialog By default DPL initializes the node as a normal distribution with the original value 0 35 as the mean You need to supply the standard deviation gt Click the edit box next to stddev Delete 1 Type 0 04 The node definition dialog should look like Figure 5 18 gt Click OK Node Definition Gross Margin General Conditioning Print Page Setup Copy FullScreen whee Reorder QQA eA Distribution type normal v Links mean 0 35 vif adder vf A Gross Margin normal Cci Hep Figure 5 18 Node Definition Data for Gross Margin Now that your model has uncertainty you can run a Monte Carlo simulation In the Monte Carlo Samples section of the Home Run group see Figure 5 19 change the initial number of samples to 100 gt Make sure Risk Profile is checked gt Uncheck all other outputs
394. s Rainbow Diagram feature allows you to run a model for several settings of an input variable Since you earlier linked the value Loss of Synergy to your model you can use this feature now Select Home Sensitivity Rainbow Diagram DPL displays the Run Rainbow Diagram dialog gt Click Select DPL displays the Select Value for Rainbow Diagram dialog Figure 5 33 gt Choose Loss_of_Synergy from the Value for sensitivity drop list Click OK Rainbow Diagram Decision Analysis Variable Variable name Value for sensitivity Loss_of Synergy ha Current value Cancel Figure 5 33 Select Value for Rainbow Diagram dialog 1 6 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models You would like to know how low Loss of Synergy would have to be in order for spinning off Enlightify to be the optimal alternative gt Type 0 03 for From gt Type 0 06 for To Type 0 01 for the Step size While you could run the sensitivity analysis with 1000 samples the randomness of the simulation draw might obscure the effect of changing the Loss of Synergy value For better results increase the sample size to 5000 Click Change The Run Options dialog appears gt Change the Initial number of samples to 5000 gt Click OK The Run Rainbow Diagram dialog should look like Figure 5 34 Rainbow Diagram Decision Analysis Venable Vanable name Loss_of Synergy S
395. s Results Syncopation Software Conditioning Sales 2 The probabilities for this node depend on Cost per Unit Marketing Costs Price Price Change Product Strategy Number of probability expressions required 9 Cancel Figure 3 19 Conditioning Dialog for Sales 2 Probabilities As the Node Definition probability tab indicated nine data expressions are required The probability input tree for this is rather small Click Full Screen to make entering the data easier See Figure 3 20 Enter probabilities for Sales 2 according to Table 3 7 Sales Outcome Sales 2 Outcome Probability Low i Coy a ooma o Table 3 7 Probability Data for Sales 2 98 Syncopation Software Sales 1 Nominal High Chapter 3 Decision Analysis Results Node Dehat Sates 2 Low 6 Nominal 3 High 1 Low 3 Nominal 4 High Low 1 Nominal 2 High 7 Figure 3 20 Full Screen Probability Input Tree for Sales 2 Note that the probabilities for Sales 2 given that Sales 1 is Nominal are unchanged The probability data in Table 3 7 reflects the belief that if sales in the first period are low i e if the outcome of the Sales 1 chance node is Low then sales in period two are more likely to be low i e the outcome of the Sales 2 chance node is more likely to be Low Similarly if sales in the first period are high they are more likely to be high in the second period Click Full
396. s the keywords representing states of another event if you specify states for that event both earlier and later in the sequence section DPL uses the columns in which these keywords occur to distinguish the states of one event from those of another If there is any ambiguity about which event is associated with an event state DPL associates the state with the event that appeared most recently in the program The following sequence section is incorrect sequence incorrect code decide to Get Into Yes and pay Info Cost then gamble on Market Forecast then decide to Release Product Yes and gamble on Demand and get Profit to Release Product No to Get Info No then decide LO Release Product Yes and gamble on Demand and get Profit to Release Product No The keywords for Get_Info and Release_Product should not start in the same column The following is correct 415 Chapter 15 DPL Programs and Code Syncopation Software sequence 7 7 Correct gode decide to Get Info Yes and pay Info Cost then gamble on Market Forecast then decide to Release Product Yes and gamble on Demand and get Profit to Release Product No to Get Intro Nor then decide to Release Product Yes and gamble on Demand and get Profit to Release Product No 15 5 4 Duplicating Parts of a Tree Performing a Subtree You can duplicate repeated sections of the sequence section with a perform link You can define a perform link by defining labels in the sequence section
397. s the only tornado analysis available gt Value Tornado is the default tornado type indicated by the Tornado split button icon To run a Value Tornado click Home Sensitivity Tornado or press F8 DPL displays the Model Get Pay dialog see Figure 5 11 This gives you an opportunity to tell DPL which value you would like to calculate In your model Total Value is the only import node and hence the only output of the spreadsheet and is the only reasonable choice so DPL suggests that in the Model Get Pay dialog Enter the variable or expression that you would like OPL to calculate as the final output e g Profit NPY Total Cost Total Value Cancel Figure 5 11 Model Get Pay Dialog gt Click OK to accept the default Total_Value Next DPL displays the Value Tornado setup dialog 152 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models Value Tornado Setup Modell Value i Current Bad_Debt_Allowance 0 03 Gross_Margin 0 35 Other_Cost_Growth_Rate 0 1 RMC_Lawsuit_Settlerment 65 Share_of_Revenue_at_Risk 0 1 Share_of_at_Risk_Revenue_Lost 0 7 Trendline_Revenue_Growth_Rate 0 08 Working_Capital 120 Remove Value Current value Bad_Debt_Allowance 0 03 add Replace Conditioning Figure 5 12 Value Tornado Setup Dialog The Value Tornado Setup dialog provides a table in which to enter the low and high settings range for each value node that is a constant
398. s tutorial but you should note the following components of the model If any of these components are unfamiliar you may wish to review Chapter 2 3 4 and or 6 of this manual Specifically this model contains the following e Several chance nodes some of which are conditioned by other chance nodes see Chapters 2 and 3 Not all the nodes are linked to the Excel cash flow model e Several value nodes which serve various purposes import nodes export nodes and parameters p and q that can be used for sensitivity analysis see Chapter 3 e A strategy table comprising two decisions pipeline capacity and whether to expand the pipeline project to a second riskier market see Chapter 4 e Asymmetry for example the Market 2 Political Stability Future node only matters if the project is expanded to Market 2 see Chapter 6 255 Chapter 9 Multidimensional Value Node Syncopation Software Right now the model has two Excel linked import nodes Upfront_NPV and Total_NPV You will look at the Excel links and add another import node a value node for the 10 year annual cash flow Launch Excel gt Open the spreadsheet called Multidimensional Value xls Navigate to the Examples folder where you installed DPL The default location path is C Program Files x86 Syncopation DPL8 Examples Select the Assumptions sheet Figure 9 2 x WW Jr ls Multidimensional Value xls Compatibility Mode Microsoft Ex
399. s within the range determined by the horizontal position and width of the bar For example in the frequency histogram Risk Profile chart in Figure 3 8 there is approximately a 4 8 chance that the objective function falls in the range 148 to 152 the height of the highest bar is 4 8 on the y axis it starts at 148 on the x axis and ends at 152 A Risk Profile chart can be formatted in a number of ways If you like experiment with some of the other formatting options available on the Format tab Figure 3 9 demonstrates some of the ways in which a chart can be formatted If you d like your chart to match that in Figure 3 9 do the following Revert back to a Cumulative risk profile chart type by un checking the Histogram checkbox in Format Display Create a title of Expected Value for the chart by checking the checkbox next to Chart in the Format Title group To rename the title double click the default title Chart 1 and type in Expected Value 88 Syncopation Software Chapter 3 Decision Analysis Results cy Select the Chart title text and change the font within View Font You can do the same with the values along both the x and y axis and the axis label Cumulative Probability Within the Format Display group uncheck Shading This will cause the background shading of the chart to go away Uncheck the checkbox next to Color Fill within the Format Color group This will leave a single line with no color fil
400. se Case Tornado Diagram Probabilistic Base Case Tornado Diagram Initial Decision Alternatives Tornado Diagram and Event Tornado Diagram Separable Value Model A model whose value function contains terms that are combined together by addition and subtraction It is possible to promote terms in these models which may allow for faster calculations Sequence Section Systematic English language descriptions of decision tree structures written in DPL code The Sequence Section tells DPL in what order to evaluate nodes and get pay expressions Series A type of variable that allows you to define a one dimensional set of values formula using intervals Subsequent elements of a series can depend upon earlier elements Similar to a a set of formulas in a row in Excel where column n 1 depends upon column n Session Log A pane located within the Workspace Window which maintains a record of the DPL session Error messages and DPL commands performed in the Model Window will be automatically written to the Session Log Information from an analysis can also be written to the Session Log see Display Function 474 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms Spreadsheet Model A spreadsheet which contains a set of calculations which evaluates one scenario of a decision analysis DPL links to the spreadsheet model in order to rapidly evaluate a wide range of scenarios based on specified levels of uncertainty Spr
401. seminars can be provided as well Please contact us at sales syncopation com if you are interested in DPL training Syncopation Software Chapter 1 Introduction 1 3 INSTALLATION If you have not already installed DPL do so now gt If you received your DPL 8 software and license electronically you should have received an e mail with the subject DPL download instructions from sales syncopation com Follow the instructions contained in the e mail gt If you purchased a DPL 8 installation CD insert it into your CD DVD drive If the setup program does not launch automatically double click on DPL8 msi in the root directory of the CD from Windows Explorer Once the DPL8 msi file has been opened follow the on screen instructions 1 4 A BRIEF TOUR OF DPL 1 4 1 The DPL Workspace Start DPL by double clicking the icon on your desktop az or double clicking the application in Windows Explorer DPL will load with an empty Model Window maximized on the right hand side of the screen On the left hand side will be the Workspace Window See Figure 1 1 Chapter 1 Introduction Syncopation Software Gd DPL Workspace Model oa via Heit rg wa 6 gt c 3 sa Eat Change Aad Ada From foe Sormeta _Opy Peroetual Figure 1 1 New Workspace with Empty Model Window 1 4 2 The Workspace Window The Workspace Window is divided into two different panes 1 the Workspace Manager
402. single interval 10 element series with with no data in it gt Click on the Links tab This node like the other linked nodes in the model is linked to Time Series Percentiles xls It is an import node since it has no data See Figure 10 2 is Node Definition Ten Year Cash Flow General Calculation links These links are will be recalculated throughout the run They may be either inputs used in the spreadsheet calculations g unit sales or calculated results from Data the spreadsheet e g NPV None local DPL Program Microsoft Excel Workbook C Program Files 86 Syncopation DPL8 Examples Time Series Percentiles xls Browse Sheet Cell Assumptions Ten_year_CF Go to Cell Cell Names y This is an import node because it doesnt have data lts value will be recalculated by Excel throughout Paste Link the run Initialization links These links are for connecting to a spreadsheet or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at the start of each run None local or DPL program Microsoft Exce Import nodes dont need to be initialized Figure 10 2 Links for Ten Year Cash Flow Node gt Click Cancel Time Series Percentiles use DPL s attributes interface The next step to run Time Series Percentiles is to define attributes for each time peri
403. sion Analysis Models Note that the Low value for Cost per Unit is the same as the current value Assume you have reliable information that the cost cannot be less than 0 8 You have now selected the values to include in the Value Tornado and specified ranges for each Click OK to run the Value Tornado DPL generates the Value Tornado as displayed in Figure 2 29 Juna s aa DPL New Preduct Strategy da Teenadet oEN Relie Value Tornado f xpa ted Value BD Low Policy Change I Policy same as Nominal E High Policy Change gt Lente vt in house X Terma x 1p Figure 2 29 Value Tornado Diagram The x axis horizontal axis of the Value Tornado diagram displays the change in the objective function of the model as each of the variables on the left is changed from the low setting you specified to the high setting you specified For a given variable e g Unit Sales it is changed from the low setting to the high setting while all other variables are held at their current settings DPL runs the model once to establish the value of the objective function with all variables at their current setting for a model with only value and decision nodes This is called the Base Result and is indicated by the vertical line in the Value Tornado See Section 2 7 for a further discussion of the Base Result Note that the vertical line on the diagram is at 105 1 which is the value you obtained running the model previously DPL sorts the bars so
404. sion cell for the CF_2014 attribute the second cell gt Click the variable button x Select Ten_Year_Cash_Flow from the list of variables gt Click OK to return to the Branch Definition dialog 284 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics The Get Pay for CF_2014 is now defined as Ten_Year_Cash_Flow but as you know this variable is a 10 element series You want the first element Ten_Year_Cash_Flow 0 to be used for the attribute CF_2014 You can have DPL do this for you and set up the expressions for the remaining attributes with one button click gt Click the Fill Down button ta DPL initializes the second attribute and the remaining nine attributes to be the elements of the Ten_Year_Cash_Flow value node Ten_Year_Cash_Flow 0 through Ten_Year_Cash_Flow 9 The dialog should now look like Figure 10 8 Branch Definition Get Pay CapEx Get or pay expressions Control esa a E sik el oa NPV i Total NPV isplay O E ee LU ce f ii CF 2014 Ten Year Cash Flowdlll CF 2015 Ten_Year_Cash_Flow 1 CF_201 amp Ten_Year_Cash_ Flow 2 CF_2017 Ten_Year_Cash_Flow 3 CF 201 Ten_Year_Cash_Flow 4 CF 2019 Ten_Year_Cash_Flow 5 CF _2020 Ten_Year_Cash_Flow 6 CF_2021 Ten_Year_Cash_Flow 7 CF 2022 Ten_Year_Cash_Flow 8 CF 2023 Ten_Year_Cash_Flow 9 Go to Attributes Ga O Pay LI yif OK Cancel Help Figure 10 8 Get Pay tab of Branch Definition Dialog with New Ge
405. sion optimization Series optimization Lottery optimization Risk Profile Number of intervals 500 Default Type Intervals all samples also used in Time Series Percentiles Cumulative Use selection from Policy Tree for Scenario Risk Profile settings Histogram Other outputs Dont display in grid Display Prompt Gather attribute expected values in the Policy Tree multiple attribute models orby Wam if large number of Rainbow Diagram steps Use default states defined in node definition for Option Value Chart Figure 10 11 Run Options Dialog 289 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software Risk Profiles underlie the Time Series Percentiles output In the Run Options dialog you can change the number of intervals to be used to accumulate Risk Profiles for the Time Series Percentiles chart This setting is the number of bins into which the outcome values will be grouped The accuracy of the percentiles depends on the number of intervals setting The more intervals you use the greater the accuracy however more intervals requires more memory and will have an impact on model runtime Usually 500 intervals should suffice If you have extremely long tails in your distributions you may wish to increase the number Note the number of intervals setting is also used for Risk Profiles in a Decision Analysis run If you change the setting here your new setting will also be used the next time you run a Decisio
406. sn t practical for DPL to implement every Excel function If you know in advance that the spreadsheet you are building will be linked to DPL you can follow the guidelines in Table 16 4 to ease spreadsheet conversion Be sure to contact Syncopation technical support if you have questions about these suggestions or other conversion issues It is not uncommon for experienced DPL users to need a little help converting very large or complex Excel spreadsheets Spreadsheet Suggestion modeling issue Table lookups Use VLOOKUP HLOOKUP INDEX Don t use OFFSET Dates Use numerical calculations Excel date functions are not supported Iteration Spreadsheets that iterate cannot be converted and tend to be slow even when linked Iteration is often used for recursive calculations such as interest which can be solved explicitly Rounding Don t use ROUNDUP ROUNDDOWN use ROUND Miscellaneous Use array formulas in place of SUMIF COUNTIF RANK Table 16 4 Spreadsheet Modeling Suggestions for Easier Conversion 434 Syncopation Software Chapter 16 Converting Spreadsheets to DPL Code While following these suggestions may represent a significant change of practice keep in mind that a converted spreadsheet often runs 10 100 times as fast as a linked one so for a large model it can be worth the effort Syncopation sometimes makes enhancements to DPL s spreadsheet conversion between major releases Check our website or contact technic
407. specified probabilities DPL evaluates each chance outcome to calculate the expected value of the decision model when rolling back the decision tree A discrete chance node is represented in the Influence Diagram by a bright green oval You will now change the value nodes that are decision sensitive into discrete chance nodes Select Unit Sales gt Click the drop down split of the Model Node Change To button on the ribbon see Table 1 2 and select Discrete Chance from the list gt The Node Definition dialog appears as shown in Figure 2 31 Node Definition Unit Sales Generat Name and settings Data Unit __ Hide Sales link _ Separate probability and value data input trees inks DPL variable name for expressions Unit_Sales Outcomes Low Add Nominal Default Delete Rename F2 Default Comments Print All Comments Copy to Clipboard ox Canes Ha Figure 2 31 Node Definition Dialog General Tab for a Discrete Chance Node 59 Chapter 2 Building Decision Analysis Models Syncopation Software The General tab of the Node Definition dialog for a discrete chance node contains an Outcomes list box You can add delete insert or rename chance outcomes DPL has created three chance outcomes and named them Low Nominal and High These defaults can be changed in File Options General Note As you saw with decision node alternatives in Section 2 4 DPL has designat
408. st interval is for the first year and the second is for every year thereafter The intervals in a series do not have to start at 0 as DPL arrays do Your series will use the cash flow years for the subscripts for clearer illustration You will also see how to use a relative subscript in this example gt In the From cell for interval 0 type 2014 gt In the From cell for interval 1 type 2015 Note that DPL fills in the To cell for the preceding interval so that the interval subscripts are consecutive In the To cell for interval 1 type 2038 Note you must specify the To boundary for the last interval Click in the Expression cell for interval 0 Click the Select Variable w button Double click on Market_1_ Size in the list a g 267 Chapter 9 Multidimensional Value Node Syncopation Software Click in the Expression cell for interval 1 Click the Select Variable M button again Double click on Mkt_1_All_Years in the list In the Expression cell for interval 1 edit the expression to read Mkt_1_All_Years 1 1 Mkt_1_Growth Note you can use the Select Variable My button again if you wish to bring in the Mkt_1_Growth variable The dialog should now look like Figure 9 13 GU Node Definition Mkt 1 All Years General a A Intervals 2 Series interval entry Create M x 1 column vector ea vi f Interval From To Expression 0 2014 0 1 2015 2038 Mkt_1_All_Years 1 1
409. strategy node Subscript Used to reference particular elements of an array or series For a series a subscript ranges between the lower and upper boundaries of the series For using subscripts with arrays see Array Subscripts Subtree A section of decision tree from a given node in the tree forward i e to the right Symmetric Node A node in a decision tree whose branches all lead to the same subsequent event Symmetric Tree A decision tree in which every node is symmetric 4 5 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software T Time Series Percentiles A graph that displays the range of outcomes over several time periods e g yearly cash flows over a 10 year period Tolerance See Risk Tolerance Coefficient Tornado Diagram A sensitivity analysis that displays the value and policy impacts of varying input values See Value Tornado Diagram Base Case Tornado Diagram Probabilistic Base Case Tornado Diagram Initial Decision Alternatives Tornado Diagram and Event Tornado Diagram Triangular Distribution A named distribution supported by DPL For a detailed description see On Line Help Two Way Rainbow Diagram An analysis tool which varies two values over user specified ranges and displays the impact on the expected value or certain equivalent and changes to the optimal decision policy U Uncertainty See Chance Nodes Uniform Distribution A named distribution supported by DPL
410. t Pay Expressions 285 Chapter 10 Time Series Percentiles and Multiple Metrics Syncopation Software More information is given in Section 10 3 on how to use the Fill Down button Click OK You can see in the Decision Tree pane that DPL has added the ten new attributes to the Get Pay expression on the CapEx node of the decision tree This causes the branches of the node to be quite long Select the branches of the CapEx node in the Decision Tree pane gt In the Model Get Pay group select the Indicator radio button Note that you could alternatively choose to show nothing by selecting the Hide radio button The Get Pay expression is no longer displayed on the CapEx node Next you will repeat this process for the other tree branch that has a Get Pay expression the No branch of the Proceed decision gt First select the No branch of the Proceed decision node in the decision tree and again select the Indicator radio button in the Model Get Pay group This will prevent the new actual Get Pay expressions from being shown so you won t have to reformat the tree later gt Repeat the above process to add the Ten_Year_Cash_Flow value node to the Get Pay expression defined for the No branch of the Proceed decision Note that the value for the NPV attribute for this branch is Upfront_NPV See Figure 10 9 286 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics Branch Definition Ge
411. t Pay Proceed Get or pay expressions Control ee Esa NO oT al di li NPV Upfront_NPV CF 2014 Ten_Year_Cash_Flow 0 CF 201 Ten_Year_Cash_Flow 1 CF_2016 Ten_Year_Cash_Flow 2 CF_201F Ten_Year_Cash_Flow 3 CF_201 amp Ten_Year_Cash_Flow 4 CF_2019 Ten_Year Cash Flow 5 CF_2020 Ten_Year_Cash_Flow 6 CF 2021 Ten_Year_Cash_Flow 7 CF_2022 Ten_Year_Cash_Flow 8 CF 2023 Ten_Year_Cash_Flow 9 Go to Attributes Ge O Pay 1 V Figure 10 9 Get Paytab of Branch Definition Dialog for Proceed No with New Get Pay Expressions Note that the same value node is being used for the ten years of cash flow in both Get Pay expressions This works because the logic is set up in the Excel spreadsheet so that these ten annual cash flow values reflect whether the Proceed decision is Yes or No The model was set up this way for ease of illustration In other situations the spreadsheet might have been set up differently so that an entirely different value node or group of value nodes would be used for each Get Pay in the decision tree Note you can copy and paste Get Pay expressions from one branch to another There are two ways to this If you wish to copy the entire Get Pay expression for all attributes or if it is a single attribute model select the branch whose Get Pay you d like to copy and press Ctrl C and then select the branch where you would like to paste the Get Pay and press Ctrl V If you d like to copy a subset
412. t does not display easily on a single screen You can change what is displayed for Get Pay expressions in the Decision Tree gt Double click the Wildcat item in the Workspace Manager to activate the model you were just working on Select the branches of the Oil Price node in the decision tree gt Select the Indicator radio button in the Model Get Pay group See Figure 8 14 248 Syncopation Software Chapter 8 Objective Function and Multiple Attributes Profit Revenues Vy F j Show Hide G Indicator Get Pay Figure 8 14 Get Pay Display Options in Model Get Pay Group You could also change the Get Pay display settings within the Display tab of the Branch Definition dialog Model Get Pay Options Display Directly on the ribbon or within the Branch Definition dialog you can choose to display the actual get pay the default display an indicator or nothing Additionally within the Display tab of the Branch Definition dialog you can hide or display state names in the decision tree One last display option for instances in the decision tree can be accessed by double clicking the node in the decision tree to open the Tree Instance tab of the Node Definition dialog Under Display options you can show hide underline and or put a border around the node name Your Decision Tree should now look like Figure 8 15 Drilling Costs Med Ww Nominal lt d 0 Land_Disturbed Drilling_Costs 0
413. t is selected as shown in Figure 4 8 130 Syncopation Software Chapter 4 Strategy Tables a Node Definition Strategy General Commands Print Page Setup Copy Format Full Screen Dat Feature Feature Feature Marketing SetB Set C Spend Add features Add features Increase Increase Increase Increase a a Keep same Keep same Keep same Keep Same Keep same Keep same Keep same Minimal Minimal Minimal Decrease Decrease Decrease Decrease Figure 4 8 Selection in the Strategy Window Right mouse click to bring up the Strategy Window context menu Select Rename DPL enters edit mode as shown in Figure 4 9 a Node Definition Strategy General Commands Print Page Setup Copy Format Full Screen Seale naw ame awe Marketing Spend Strategy Addfeatures Addfeatures Addfeatures Increase Increase Increase Increase Bl strategy2 Keepsame Keepsame Keepsame KeepSame Keepsame Keepsame Keepsame Minimal Minimal Minimal Decrease Decrease Decrease Decrease Figure 4 9 Edit Mode in the Strategy Window 131 Chapter 4 Strategy Tables Syncopation Software Rename the strategy to Status Quo Press Enter The Strategy Table Window should look like Figure 4 10 a Node Definition Strategy General Commands Print Page Setup Copy Format Full Screen Data i CI Feature Feature Feature Marketing Price
414. tab Type 0 5 as the value gt Click OK Your Influence Diagram should look something like Figure 8 5 You ve made the required changes to the Influence Diagram You ll now review the Decision Tree 259 Chapter 8 Objective Functions and Multiple Attributes Syncopation Software Figure 8 5 Influence Diagram with New Nodes Press the Tab key to activate the Decision Tree The Decision Tree in Figure 8 6 has three Get Pay expressions one for each of Test_Costs Drilling_Costs and Revenues At this point all three of these expressions contribute their value to a single implicit attribute which if it were named would probably be called Profit Drilling Costs Drilling_Costs Med Drilling_Costs High Drilling_Costs Seismic Structure No Core sample Test_Costs Closed Oil Price Revenues Wet Nominal Revenues d Soaking High Revenues Figure 8 6 Decision Tree for the Single Attribute Model 240 Syncopation Software Chapter 8 Objective Function and Multiple Attributes Next you ll add the attribute for Land LUUL Gee geese eee 4 Click Model Objective amp Utility Options DPL displays the Objective tab of the Objective amp Utility dialog Click the Rename button to rename the first attribute Profit Click Add to add a second attribute Rename the second attribute Land Click in the Objective function edit box Click the Attribute button i Select
415. tal Cost NPY vii Cancel Figure 2 12 Model Get Pay Dialog Click OK to close the Model Get Pay dialog and run the analysis DPL runs the model and brings up the Analysis Complete dialog See Figure 2 13 Analysis Complete Expected value 105 1 Figure 2 13 Analysis Complete Dialog You will not normally see the Analysis Complete dialog since it is only displayed if no other outputs are requested in Decision Analysis options within the Home Run group In this instance the model is deterministic it does not yet include any uncertainty i e it contains no chance nodes which you will learn about in Section 2 6 so the expected value is just the calculated value of NPV gt Click OK to dismiss the Analysis Complete dialog 36 Syncopation Software Chapter 2 Building Decision Analysis Models 2 4 DECISION NODES You will now enhance your DPL model by adding a decision node A decision node has two or more alternatives DPL will choose the decision alternative that maximizes the output of the model referred to as the objective function A decision node is represented in the Influence Diagram by a yellow rectangle Note that the default add node type currently indicated by the icon in the Model Node Add split button see Table 1 2 is value To create a decision node drop down the Model Node Add split button and select Decision from the list The default add node type changes based on th
416. tal number of nodes is also given The linked spreadsheet in Figure A 14 has 14 calculation links nodes and 2 initialization links nodes out of 14 nodes in the model A couple of the nodes in this model have both calculation links and initialization links After you ve set up the links in your model it s a good idea to use Show Nodes to check that all the nodes are linked to the correct cells see Figure A 14 This is particularly true if you used Method 3 to establish the links since you could have chosen the wrong cell by mistake The dialog shows each linked node the sheet and range name to which it is linked the link type and node type A node is shown twice if it is both calculation linked and initialization linked 453 Appendix A Overview of Spreadsheet Linking Syncopation Software Link Type Node Type Initial license Assumptions amp Output lnitia i Discrete Chance Initial training Assumptions amp Output Initial Discrete Chance License phase 1 growth Assumptions amp Output Licen Discrete Chance Marketing phase 1 Assumptions amp Output Mark Discrete Chance Marketing phase 1 probs Initialization Mkt_ph1_values Discrete Chance Marketing phase 2 Assumptions amp Output Mark Discrete Chance Marketing phase 2 probs Initialization Mkt_ph2_cond_p Int Discrete Chance C Show unlinked Training phase 1 growth Assumptions amp Output Traini Discrete Chance nodes Consulting phase 1 growth Assump
417. tcomes for a chance event and the number of elements in a series Unlike values named numeric constants cannot depend on events Once you define a named numeric constant you cannot change it with a DPL command or perform a sensitivity analysis with respect to it However you can change the definition of a named numeric constant in a DPL program file and then recompile the program This allows you to consistently and easily change parameters of a decision analysis throughout a DPL program 15 4 13 The Integer Keyword The DPL keyword integer can be used in the definition of constants values series and arrays that must have integer values Whenever DPL evaluates an expression that will be assigned to an integer variable the expression result will be rounded to the nearest integer value For example integer value Years Remaining years 1 83 integer Years Remaining years 1 83 same as above integer series index from 2014 1 from 2011 to 2016 2 6 factor For most purposes the integer keyword and the round function have the same effect the integer keyword is just a convenience For example if you want all the values in an array to be integers you can use the round function in the definition of each array item or you can define the array to be an integer array and the rounding will be automatic If you have a variable that must be an integer and you are using the dont specification to remove lotteries from th
418. ter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Node Definition Gross Margin General Conditioning Print Page Setup Copy Full Screen Wheel Reorder Q Q Q 4 Data 0 35 IDIF Links Gross Margin 0 35 F Cant Heb Figure 5 8 Node Definition Data for Gross Margin In Figure 5 8 you can see that DPL has brought over the data 0 35 or 35 from the Excel cell Right now this may seem like a lot of space for one number but you ll be using it as you start to incorporate uncertainty into the model Click on the Links tab In Figure 5 9 you can see that DPL has recorded the path file name sheet name and cell name for the Gross Margin node DPL also tells you that this is an export node When spreadsheet inputs or drivers are linked to DPL nodes they are always DPL export nodes whereas spreadsheet outputs e g NPV are DPL import nodes Export nodes are typically linked to cells containing data whereas import nodes are typically linked to cells containing formulas If you are unsure whether a DPL node is linked to the correct cell you can have DPL activate the cell in Excel 148 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models General Data m Node Definition Gross Margin Calculation links These links are recalculated throughout the run They may be either inputs used in the spreadsheet
419. tes a chart showing the expected value of perfect information and the expected value of control of each discrete chance node in the model Tells DPL how many levels down the tree to gather policy information A tree with N nodes in its longest path will have N 1 levels Must be set to all for Expected Value of Perfect Information Control Tells DPL to record an endpoint database for the run The endpoint database allows you to replay the endpoints to quickly perform further analyses without recalculation If Endpoints is checked the evaluation method is set to full tree enumeration 73 Chapter 3 Decision Analysis Results Syncopation Software Value Correlations Creates a bar chart showing the correlation coefficient between each value in the model and the objective function Table 3 3 Policy Output Options The last section of the Home Run group contains two buttons Clear Mem and Compile Clear Mem tells DPL to clear the compile structures and any unsaved results from a Decision Analysis run for the active model The Compile command tells DPL to compile the active model to test for missing data or other errors Note that there are two additional evaluation methods in the Decision Analysis drop down list not included in Table 3 1 Monte Carlo simulation and Full Tree Enumeration from Endpoints Monte Carlo simulation and the options associated with this method will be covered in Chapter 5 Full Tree Enumerat
420. th an expression The expression can contain values that depend on events and any of DPL s functions DPL allows you to reference an array before defining it This makes it possible to define several arrays recursively Because you can define both series and arrays recursively you must first specify whether the variable being referenced before its definition is a series or an array if you don t DPL will assume it is a series Then use the variable name and finally provide the complete definition For example array a series s from 1 a 0 0 from 2 to 4 a l 2 array a sum s 2 4 0 0 Tas ae You can refer to arrays in three ways Assume the series Costs is defined as below array Costs 10000 20000 30000 40000 12500 22500 32500 42500 15000 25000 35000 45000 e With subscripts referring to a single element e g value YeariCost Costs 0 0 Year1Cost is 10000 e Without subscripts referring to all the elements in the array e g value CostTotal sum Costs CostTotal is 330000 409 Chapter 15 DPL Programs and Code Syncopation Software e With a subrange specification referring to a selected subset of the elements in the array e g value CostSubTotal sum Costs 0 1 1 3 CostSubTotal is 187500 15 4 11 Strings You can define a variable called a string and initialize it with a sequence of characters string Report Title Fuel Price Analysis Strings can depend
421. that has two elements corresponding to years 2015 and 2016 and one that has four elements corresponding to years 2017 2018 2019 2020 A series has at least one interval definition that contains an expression that gives the number represented by the series in that interval The interval definitions also contain the interval boundaries in this case 2014 2015 2017 and 2020 The lower bound of the first interval and the upper bound of the last interval define the lower and upper bounds of the series and they must be constant expressions The other interval boundaries are called intermediate interval boundaries and they may be variable expressions Only the upper bound for the last interval must be explicitly defined the upper bounds for the other intervals are calculated by subtracting one from the lower bound of the next interval The interval bounds and subscripts must be integers and they can cover any range that you designate including negative numbers Each interval boundary must exceed the previous one All interval ranges must lie between 32768 and 32767 If necessary DPL will truncate subscripts and interval bounds to integers A series also has a subscript a way to reference particular elements of the series which can range between the lower and upper boundaries of the series In this example the subscript may range from 2014 to 2020 A basic subscript refers directly to a particular element in the series For example
422. the punctuation in the examples as a guide DPL will usually catch punctuation errors during compilation If it does find such an error it will return you to the error location DPL doesn t care how many lines a definition covers or how many spaces or tabs you include between the words in a definition The only restriction on spacing is that identifiers the names of events states values series and so on must not include blanks or tabs Everywhere else in the definition section you may space things however you like Note that this is 399 Chapter 15 DPL Programs and Code Syncopation Software not the case in the sequence section if the model is asymmetric see Section 15 5 15 4 3 Excel Linked Events and Values Values and events linked to Excel will be preceded by an excel clause as in the following excel Excel 1 Assumptions Market 1 Size value Market 1 Size 110 The excel clause shows the workbook path Excel_1 and the sheet and cell Assumptions Market_1_Size to which the value or event is linked In most cases all nodes will be linked to the same workbook and the path and file name of that workbook will be defined as a string named Excel_1 If the Excel linked value is an import node then there will be no assignment following the value as in the Total_NPV example above 15 4 4 Decisions The following statement defines a decision in DPL decision Research Funding Low Medium High The tilde
423. the links as a group using the Model Links dialog by selecting Model Links Options See Figure A 13 From this dialog you can Change the spreadsheet name or location Remove all links to the spreadsheet Browse the nodes linked to the spreadsheet Create a link to a new spreadsheet Convert the spreadsheet to a DPL Calculation Program DPL Professional and Enterprise only 452 Syncopation Software Appendix A Overview of Spreadsheet Linking Model Links Cash Flow with Decision Extemal Links Settings Spreadsheet Database Name i Calc Init Nodes Cash How with Decision xs Esce 14 2 of 14 Browse Remove Show Nodes Convert to Cale Program Create Link to New Spreadsheet Links to Programs in the Workspace DPL Program for Data Definitions DPL Calculation Program Values defined in this m This position is normally used for converted aes program can be used in Bisi spreadsheets Values defined in this program N both the influence diagram ER can be used as calculation links or in the eskuli and the decision tree piketai decision tree Update Update Revert to Linked yde OK Cancel Help Figure A 13 Model Links Dialog The Model Links dialog also tells you how many nodes are linked to the spreadsheet The number of nodes with calculation links is displayed first in the Nodes column of the Linked Spreadsheets list The number of nodes with initialization links is displayed second and the to
424. the objective function Value Correlations Charts help you determine which variables may be driving your objective function in either a positive or negative direction Value Model A model that defines a particular value of interest e g costs profits social welfare from other data A DCF discounted cashflow valuation spreadsheet is an example of a value model Value Node A node in the Influence Diagram representing a number or an expression Its graphical symbol is a blue rounded rectangle Value of Control The difference between the expected value of the model and the expected value of the model when a particular uncertainty is changed into a decision Value of Perfect Information The difference between the expected value of the model and the expected value of the model when the outcome of a particular uncertainty is known before a decision is made Value Tornado Diagram Evaluates the outcome of a model when a selected set of variables are each set individually to high and low values while all other variables are treated in the way in which they are defined in the base model If the model is deterministic than all other variables are held at their current value If the model is probabilistic then all other variables are allowed to take on their full range The low and high results for each variable are compared to the results of the full model run called the Base Result When used on a deterministic model the diagram is used to
425. the thumb tack button on the top right of the pane When the thumb tack is in Pin Down mode the pane stays visible Pressing the thumb tack button will toggle to Pin Up mode and the pane will collapse to a tab on the edge of the screen if docked or to a window bar if un docked When you mouse over the tab or window bar the pane will expand to its visible state When your mouse leaves the tab pane area the pane will collapse to a tab window bar again To keep the pane at its visible state click the thumb tack to toggle to Pin Down mode 1 4 3 The Model Window The Model Window is also divided into two different panes 1 the Influence Diagram pane at the top and 2 the Decision Tree pane at the bottom The two panes are divided horizontally by a splitter which can be moved up or down Dragging the splitter bar to the edge of the window will make only one of the panes visible as will double clicking on the splitter bar Double clicking the splitter bar again sets it to the half way split To change back and forth between panes press the Tab button click anywhere within the desired pane if visible or click View Window Switch Panes Decision Tree or Influence Diagram The active pane is indicated by a magenta outline around the edge of the pane You can also change the splitter to split the screen vertically by selecting the radio button View Layout Split Vertically which results in the Influence Diagram being on the left and Decisi
426. ti Percentiles a Figy Mod Session Log 7 DPL Desmar Programming Language Caprio E 20012 by Synenseion So Soe E T os Weporiiiccpai tan ES Wikies 1 5E T9 7375 Duite he Wo 1 PrE JI ee Eniemprieet Verson irierial Buki Prr pitaal Lens Meinaan 00 0G Beta 3 tert Use Cirih oo Energy Model x ab Fie Help pitii Fl foe GSN Full Figi ma Figure 10 1 Energy Case Model Workspace This model is the same as the model that is used and modified in Section 9 1 1 Alternatively if you ve already worked through that section you may use the model you saved then You also need to open the Excel model you saved or alternatively Launch Excel gt Open the spreadsheet called Time Series Percentiles xls Navigate to the Examples folder underneath where you installed DPL If you used the default location the path is C Program Files x86 Syncopation DPL8 Examples The first step to run Time Series Percentiles is to create a value node for the metrics you are interested in seeing over time This was done in Section 9 1 1 The Ten Year Cash Flow value node is already linked to the spreadsheet and set up to import 10 years of cash flow results into DPL To review this do the following Switch back to DPL gt Double click on the Ten Year Cash Flow value node to open it 2 8 Syncopation Software Chapter 10 Time Series Percentiles and Multiple Metrics gt Click on the Data tab This node is a
427. tion sssssssssssssssnssnssnnsnrnrrenrnnnnnrnnnnnnnnnennnnne 159 5 7 COMparing RISK Profiles a E a a 164 5 8 Modeling an Up Front Decision sssssssssssssssrsrrsrnsnnnnrrnnrnnrnnrenrnnrnnrnnennnnnns 169 5 9 Performing a Sensitivity Analysis ccccccceccseeseeeeeeeeeeeeeseeeaeessesaeeseesaeeneees 176 5 10 Modeling a Downstream Decision sssssssssssssrsrrsrrsrrnnrnnrrnnrnrrnnrnnrnnennnnnnnnns 179 6 Asymme trie TrecS sasirnane aiaa aN 183 6 1 Eliminating Redundant PathS ss ssssssssssssnsrnnrnsrnsnrnnrnsnrnnrnnrnnrnnrnnrnnenresnnns 183 6 2 Using Multiple Get Pay Expressions cccsscceceseeeeeeeesseeeeeesesaeeseesaeeseesaeees 196 6 3 Adding a Downstream Decision cccceceeeeeeeeeeeeeeeeaeeeaeeea ees eesaeesaeesaeesa eens 203 7 Conditioning and Learning in Decision Models sssssseseseseees 215 7 1 Incorporating Imperfect Information in a Model cccceeeseeeeeeseeeeeeseeeeeees 216 ii Syncopation Software Table of Contents 7 2 Summary of DPL Influence AMcs ccceccseeseesecseeenteseeceeeseeeetsaeeeeesaeeenegees 232 8 Objective Functions and Multiple Attributes cscsesesesesssses 235 8 1 Incorporating Multiple Attributes cccccceeeseeeeeeseeeeeeseeseeeeeeseeeeeeseesaeenass 236 8 2 USING a CONSEAINE FUNCUON tsi usmswoseresnnoimoumerononiwsvininsnseimasamabiemamnvtoroiieaeds 250 9 Multidimensional Value Nodes cccscseseesssseseseneneeenenenseeeeeeeeees
428. tion saving DPL Workspace files with large Endpoint Databases in them particularly those displayed in a report require significant disk space to save If you try to record and save databases of tens of thousands of endpoints or more you will see DPL messages warning you about the amount of disk space that will be required Extremely large Endpoint Databases i e with hundreds of thousands or millions of endpoints may be infeasible or inadvisable to record and save depending on your computer 305 Chapter 11 Using the Endpoint Database Syncopation Software Also please note that if you save your model while completing this tutorial you may see that it will take a few minutes to save and it may result in a fairly large 20 30 megabyte saved Workspace file If you want to be sure not to save a file this large you should select Home Run Clear Mem or press F9 before saving the Workspace file However this will clear the Endpoint Database and you will have to re run the model the next time you open the Workspace Other tips for using DPL s Endpoint Database with large models are given later in this chapter This tutorial assumes that you are familiar with the basics of using DPL Professional and with most of the material contained in the earlier tutorials contained in this manual 11 1 RECORDING AND VIEWING ENDPOINTS For this tutorial you will use the energy model that you used in Chapters and 1 of this manual
429. tion Software DPL runs the model and produces the requested outputs The Policy Tree will be active You could have requested a Policy Summary when you ran the Decision Analysis You can also create a Policy Summary by re playing recorded endpoints Here you will see how to create one from the Policy Tree gt With the Policy Tree active select Policy Policy Summary Add from the command ribbon DPL creates a Policy Summary as shown in Figure 12 12 Test Drill Amount of Oil se 0 Yes Core Sample 49 100 Exp Seismic 52 0 does not occur does not occur 0 0 Drilling Costs Oil Price Pipeline Costs does not occur does not occur 51 0 0 Seismic Structure Exp Seismic Test No Closed does not occur 0 Figure 12 12 Policy Summary for the Wildcat Model DPL displays the policy dependent probabilities for each state of each event in the decision model in the Policy Summary It also displays these probabilities graphically by drawing a blue bar proportionate in length to the probability on each branch As the Policy Summary indicates in just under half of the scenarios it is optimal to drill and in just over half it is 334 Syncopation Software Chapter 12 Advanced Decision Analysis Results optimal to not drill The Policy Tree contains all the information needed to create a Policy Summary and Policy Summary comparisons You will n
430. tion from that point forward in the subtree The risk tolerance custom utility function applies until another one is specified further down the subtree or until the end of the subtree is reached IZI Chapter 13 Risk Tolerance Syncopation Software 3 4 Syncopation Software Chapter 14 Sensitivity Analyses 14 SENSITIVITY ANALYSES Five types of sensitivity analyses Value Tornado Base Case Tornado Probabilistic Base Case Tornado Initial Decision Alternatives Tornado and Rainbow Diagram were discussed in Chapter 3 and other earlier chapters This chapter covers two further types of sensitivity analysis that DPL offers and provides some guidance on when it is most appropriate to use each of DPL s sensitivity analysis outputs 14 1 TWO WAY RAINBOW DIAGRAMS A Two Way Rainbow Diagram is a sensitivity analysis that is run on two selected values from a model The diagram displays the policy changes for each combination of the two values selected In addition you can use the Show Tips feature to see the change in the objective function of the model for each combination of the two values You will now run a Two Way Rainbow Diagram Select File Open Workspace Navigate to the Examples folder underneath where you installed DPL If you used the default location the path is C Program Files x86 Syncopation DPL8 Examples Select Product da DPL opens the Workspace as shown in Figure 14 1 375 Chapter 14 Sensitiv
431. tion on determining what policy change has occurred see Section 3 7 You have just discovered that several of the variables in your model are decision sensitive Assume you do not know the value of these variables for certain You may want to know what the optimal policy is when these variables are treated as uncertainties You will now change your model in order to treat these variables as uncertain quantities You re about to make significant changes to the model so you will create a copy of the model as it is and edit the copy This is good practice when you have results such as a Value Tornado which are based on a simpler version of your model and which you may later need to reproduce gt Right mouse click on the item for the model in the Workspace Manager The Workspace Manager context menu appears Select Duplicate The model is duplicated with the same name plus copy appended to it Click the item for the new model in the Workspace Manager and Press F2 to edit the name Rename it License vs In house prob Press Enter The new model is now active in the Model Window You may wish to save the workspace at this point 58 Syncopation Software Chapter 2 Building Decision Analysis Models 2 6 DISCRETE GHANCE NODES You will now introduce uncertainty into the model by changing some of the value nodes to discrete chance nodes In DPL a discrete chance node has two or more outcomes which occur with
432. tioning events chance Research Cost Low Moderate Outrageous given Research Funding Research Time Research Funding Research Time 0 6 0 2 LOow Short TOA 30133 Low Medium 10 2 0 4 Low Long 10 4 0 3 Medium Short 0 4 0 3 Medium Medium 10 2 0 4 Medium Long 0 2 0 4 High Short 0 2 0 4 High Medium 0 1 0 4 High Long 40 50 80 Low Short 45 60 85 Low Medium 40 60 80 Low Long 50 95 110 Medium Short 80 100 125 Medium Medium 85 105 130 Medium Long 90 105 130 High Short 95 120 140 High Medium 95 140 160 High Long This statement defines nine sets of probabilities one for each combination of the conditioning events states Research_Funding and Research_Time The order in which you specify the probabilities depends on the order in which you list the conditioning events after the keyword given You can substitute a vertical bar for the keyword given DPL uses the first set of probabilities when all conditioning events are in their first state The next set of probabilities applies when the last conditioning event in the list changes to its second state After associating sets of probabilities with all the states of the last conditioning event DPL varies the states of the next to last conditioning event and so on This continues until probabilities are associated with all combinations of states This method for determining the order of the va
433. tions amp Output Cons Value Consulting phase 2 growth Assumptions amp Output Cons Value Initial consulting Assumptions amp Output Initial Value License phase 2 growth Assumptions amp Output Licen Value NPV Assumptions amp Output NPV Value Training phase 2 growth Assumptions amp Output Traini Value Wages phase 1 Assumptions amp Output Wage Value Wages phase 2 Assumptions amp Output Wage Value Figure A 14 Show Linked Nodes Dialog 454 Syncopation Software Appendix B System Requirements amp Compatibility with Older Releases B SYSTEM REQUIREMENTS AND COMPATIBILITY WITH OLDER RELEASES DPL 8 is compatible with Windows XP Vista 7 and 8 In order to be able to link to spreadsheets you need to have Excel for Office 2003 2007 2010 or 2013 To use the Excel Side Interface you need Excel 2007 2010 or 2013 A minimum of 25 MB of free disk space is required For a complete install with all documentation approximately 45 MB is required A display with resolution 1280 by 800 or higher is required A larger higher resolution screen gives a better modeling experience DPL 8 is compatible with previous releases of DPL Decision Analysis Workspace Project files da from DPL 4 0 5 0 6 0 and 7 can be loaded into DPL 8 Note that da files are called Decision Analysis Workspace files in DPL 7 and 8 and are called Decision Analysis Project files in earlier versions You can import DPL Influe
434. tions tE 18 Walue AO 25 Create arrays bated on range s1Z6 Prefix node name with sheet name Figure 9 11 Range Names Dialog DPL adds the value node names it Mkt 1 Growth and initializes it with 0 05 which is the growth rate 5 from the spreadsheet The node is now an export node whose value will be sent to Excel If you wish to use a different value for Mkt 1 Growth you would now change it in DPL instead of in Excel since the value in DPL will be sent to Excel and overwrite any change made there You will use this value in the series you are about to create gt Click Model Node Add to add a new value node Place the node near the Mkt 1 Growth node you just created In the General tab name the new node Mkt 1 All Years cy Select 1 dimensional array N values or series in the Dimensions section cy Switch to the Data tab 266 Syncopation Software Chapter 9 Multidimensional Value Nodes Check the Series interval entry checkbox gt Change the number of intervals to 2 See Figure 9 12 Node Definition Mkt 1 All Years General Intervals 2 W Series interval entry Create Mx 1 column vector kal Vv Interval From To Expression 0 0 1 1 E A A AEAEE AA AAA AEEA A Cena Hee Figure 9 12 Node Definition Data for a Series with 2 Intervals Series are defined using intervals Your series has two intervals you will set it up so that the fir
435. torage requirements If you need to save a Workspace file which contains a sizeable Endpoint Database you can reduce the storage requirements by saving the Workspace file with the Endpoint Database but without the report You may remove the report by doing the following Right click on the Endpoint Database item and select Delete You are not really deleting the Endpoint Database itself you are merely deleting the report view of it 322 Syncopation Software Chapter 11 Using the Endpoint Database Alternatively you can click on the x button within its document tab in the Document Navigator If the report window is not maximized you can click on the close button m z at the top right of the window or icon A dialog will appear confirming that you would like to delete the Endpoint Database Again you are confirming deletion of the report view not the Endpoint Database itself Your Endpoint Database is still stored but the view is not The icon for the item in the Workspace Manager will change to a blue triangle with a database can behind it instead of a blue triangle with a grid behind it as shown in Figure 11 14 Also DPL will change the caption to be Endpoint Database to indicate that the database is no longer displayed in a window Workspace Manager a9 Energy Case da he Energy Model S Endpoint Database f Expected Value S Expected Value y Policy Tree Figure 11 14
436. ts from the spreadsheet e g NPV None local DPL Program Microsoft Excel Workbook Browse Sheet Cell Go to Cell Cell Names y This is a DPL local node Paste Link Initialization links These links are for connecting to a spreadsheet or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at the start of each run None local or DPL program Microsoft Excel Figure 3 15 Node Definition Links for Sales 1 gt Click OK to close the node definition dialog 94 Syncopation Software Chapter 3 Decision Analysis Results You have divided the sales in half because you are about to introduce a new node called Sales 2 To do this you are going to take advantage of DPL s node copy and paste capability By the time you are done modifying the model you will have created a two time period model Sales 1 will represent sales in the first period and Sales 2 will represent sales in the second period You have divided the values in half for Sales to reflect the fact that previously Unit Sales represented total sales gt With the Sales 1 node selected it should be magenta make a copy of it by pressing Ctrl C or clicking Home Edit Copy Paste the new node by pressing Ctrl V or clicking Home Edit Paste Select the new node and drag it away from Sales 1 Note You can also select and copy mult
437. ts used in the linked spreadsheet s calculations e g unit sales or calculated results from the Data spreadsheet e g NPV O None ocal ODPL Progam Microsoft Excel Workbook Multidimensional Value xs Browse Sheet Cell Assumptions Ten _year_CF Go to Cell CellNames This is an import node because it doesnt have data lts value will be recalculated by Excel throughout the run Initialization links These links are for connecting to a spreadsheet or program that contains data values and or probabilities which will be used in the Data tab to initialize nodes These links are refreshed once at the start of each run None local or DPL program Microsoft Excel Import nodes dont need to be initialized Figure 9 7 Node Definition Dialog with Linked Range Name 261 Chapter 9 Multidimensional Value Node Syncopation Software DPL tells you that the linked node is an import node As mentioned above DPL created an import because it found formulas in the named range and hence did not initialize the node with any data on the Data tab gt Click OK Your model should now look like Figure 9 8 You have created a 1 dimensional 1 x 10 import series value node that is linked to a 10 cell range in Excel You can see in Figure 9 8 that the value node Ten Year Cash Flow has a vector arrow on it to indicate that it is a 1 dimensional node rather than a scalar value Figure 9 8 Mo
438. tup Format Full Screen Status Quo Pl Feature Led Feature SetA Add features p Keep same Minimal Keep Same Keep same Keep same Keep same Decrease Decrease Decrease Decrease Price Price Price pom Change Change Change A B Cc HHF Increase Increase Increase Figure 4 12 Newly Created Strategy in Strategy Window Rename the strategy Price Cuts Repeat the create strategy process to create another strategy called Differentiate Products This will be color coded yellow Your Strategy Window should now look like Figure 4 13 Syncopation Software Chapter 4 Strategy Tables Node Definition Strategy General Commands Print Page Setup Copy Format Full Screen Data s Price Price Price geent je jame T Change Change Change B C E Status Quo Add features Add features Add features Increase Increase Increase Increase E HANO MANo PESO nAn SSO Haro sano P Price Cuts O Keep same Keep same Keep same Keep Same Keep same Keep same Keep same Differentiate Products Minimal Minimal Minimal Decrease Decrease Decrease Decrease Figure 4 13 Strategy Window with New Strategies Created DPL allows you to define up to 12 strategies in a strategy table For this example four strategies will suffice You will now make the four strategies distinct from one another by
439. tware on ieesse to Syncepalion Software he Ad nghis reserved SUPPOMMERSyNcopation com US toBfree 1 06 795 7375 Outside the US 1 97S 233 2509 Professional Version tpe sal Rewase amp 00 00 Serial Nurber SYOPB0130102001 Licensed User suser Syncopalen Sonwe 20 0 20 40 60 80 100 120 140 160 160 200 220 240 260 280 300 lt gt Licensevs In house x So License vs In house prob x Tomado x Tomado x Tomado3 x Policy Tree x i Policy Tree x db For Help peess Ft X 177 781 Y 100 Zoom Figure 3 7 Risk Profile Chart By default DPL creates a cumulative Risk Profile chart A cumulative Risk Profile chart displays the objective function on the x axis horizontal axis and cumulative probability on the y axis vertical axis A cumulative Risk Profile chart can be read by choosing a value on the x axis determining where the Risk Profile intersects with the vertical line drawn from the value and then drawing a horizontal line over to the y axis The value where the horizontal line meets the y axis is the probability that the objective function is less than or equal to the chosen value on the x axis For example if you choose the value 40 you see that the vertical line drawn from 40 intersects the Risk Profile at a probability value of approximately 20 Therefore the probability that the objective function is 40 or less is 20 To create a Risk Profile chart DPL calculates all the endpoint values objective function v
440. ue uncertainty strategy table or controlled event Normal Distribution Also called Gaussian Distribution A named distribution supported by DPL For a detailed description see On Line Help O Objective Function An expression defining the quantity to be optimized during the analysis Generally used to express the relationship between multiple attributes OLE Automation A Microsoft Windows protocol that allows one Windows application to control another DPL normally uses OLE Automation to communicate with Excel OLE Automation may be slower than DDE but is more robust Optimal Policy A set of decision alternatives that optimize the Objective Function Optimization Allows DPL to take advantage of special structural properties of the model to reduce computation time Option Value Chart A chart that displays the option value or incremental value added by flexibility in each downstream decision in the tree Outcome A state of a chance node or chance event that DPL evaluates during rollback by combining together the probability associated with the outcome and the value associated with the outcome to calculate its expected value 4 0 Syncopation Software Appendix D Glossary of DPL and Decision Analysis Terms P Pascal Distribution A named distribution supported by DPL For a detailed description see On Line Help Perform Subtree A technique allowing you to repeat a section of a decision tree in more than one pl
441. uence arc indicates timing or conditional dependence of nodes An arc from node A to node B means A influences B 467 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software Influence Diagram a graphical representation of the components of a decision problem decisions uncertainties and values and the relationships among them Comprised of nodes and influence arcs Influence Diagram pane A pane of the Model Window that provides a graphical interface to manipulate the Influence Diagram in a model Initial Decision A decision in a decision tree that occurs before any chance events Initial Decision Alternatives Tornado A diagram that displays a Base Case Tornado for each alternative of the first decision node in the decision tree Initial Uncertainties Chance events in a decision tree that occur before a downstream decision Interval A subset within a Series in which one or more elements are given by the same expression For example a series defining the number of days in each year from 2013 2016 would have an Interval from 2013 2015 in which each element would have a value of 365 The Interval for 2016 would be a single element Interval with a value of 366 Joint Probability The probability associated with a particular set of outcomes for multiple separate events The joint probability is calculated by taking the marginal probability for the selected outcome of each event and multiplying them to
442. un group check the checkbox next to Endpoints if it is not already Request a Risk Profile and a Policy Tree as well Leave everything else including Init Dec Alts unchecked Click Home Run Decision Analysis or press F10 to run a Decision Analysis DPL will open the linked spreadsheet if it is not already and run the model It may take a few minutes as this model has more than 32 000 endpoints and Excel has to recalculate results each time The expected value of this model is 34 9 as you can see in the Session Log or by zooming in on the Policy Tree When the model run is complete you will see an icon for an Endpoint Database in the Workspace Manager 307 Chapter 11 Using the Endpoint Database Syncopation Software Double click on this icon to display the Endpoint Database Note that the Grid tab appears and becomes active in the command ribbon See Figure 11 2 elas zE DPL Energy Case de Enpara Dwibasa a ph 4 rat a Fii a ro B art ve Cobumng giim et i 1 i i 1 PEHA Eniponi Fred ina F P Cep Aat Ro Mirka Pent fa pirat p bib gt na Eairgy Molai an M i L hiia t BCM Lin Low i Sabie ewr i ni atest 3 iYe lew Mii Ne BEMI Lom Low Gute Low doer F mastering 3 i Yer Line hiit T No DCH i Low Le i TEsbis Lies r gh a Poire Tes fi Vet lps ht He BCH I Lpa Lgs l ibi Lew iti PN tr Lees Wiri h BCh ii La Leow L dnbie law orm rn TH L Fr Wii 1 Mo DEM
443. use your mouse to find out which policy applies in each colored region if you have Show Tips turned on You turned these on earlier but you may need to check that they are on 115 Chapter 3 Decision Analysis Results Syncopation Software gt To check if Show Tips is on click File Options The Options dialog appears gt On the General tab make sure Show Tips is checked gt Click OK to close the dialog Place your mouse cursor over the blue square marker for 0 875 on the curve A policy tip will appear as indicated in Figure 3 34 Price Change Keep Same 100 Product Strategy License 100 v 2 gt U D a U D Q x i 0 8 License Fee Figure 3 34 Rainbow Diagram with Policy Tip Showing The policy tip tells you the expected value of the model given the value of the variable that the marker represents For example the policy tip in Figure 3 34 indicates that the expected value of the model is 114 when License Fee is 0 875 The policy tip also tells you the policy dependent probabilities for each decision in the model given the value of the variable that the marker represents The decision alternatives are displayed in the policy tip by using the decision node name followed by a colon followed by the decision alternative E g Product Strategy License The policy dependent probability for the alternative follows the equal sign As Figure 3 34 indicates the policy when License Fee is 0 875 is
444. value node called Test Costs Place it above Revenues le ae A A Re AY If you need more room in the window click the Zoom Out button on the status bar or drag the splitter bar between the Influence Diagram and the Decision Tree Your diagram should look like Figure 7 5 Figure 7 5 Influence Diagram with Test decision Test Costs depend on whether the wildcatter decides to test gt Double click on the Test Costs node In the Node Definition dialog click the Conditioning button gt Check the checkbox next to Test Figure 7 6 219 Chapter 7 Conditioning and Learning in Decision Models Syncopation Software Conditioning Test Costs The values for this node depend on Amount of Dil F Drill F Drilling Costs Number of value expressions required 2 Figure 7 6 Adding Conditioning Click OK DPL changes the data input tree to reflect the new conditioning You can now enter two data expressions one for each alternative of the Test decision Figure 7 7 220 Syncopation Software Chapter 7 Conditioning and Learning in Decision Models Node Definition Test Costs General Conditioning Print Page Setup Copy FullScreen Whee Reorder QQA oso G Links Test None Costs Test 0 Core Sample 6 Figure 7 7 Node Definition Data for Test Costs gt Type 0 for the None branch gt Type 6 for the Core sample branch gt
445. ve now entered a value The use of a value such as p in the probability expression of an outcome is called a non constant probability expression When you use a non constant probability expression for any of the chance outcomes in a node you must leave the last probability expression blank This helps to ensure that the probabilities of all the outcomes sum to one Click OK to close the Node Definition dialog Click Home Sensitivity Rainbow Diagram The Run Rainbow Diagram dialog appears Click the Select button Select p in the Select Value for Rainbow dialog Click OK In the Run Rainbow Diagram dialog specify the From and To values to be 0 3 and 0 9 respectively Set the Step size to be 0 05 The Run Rainbow Diagram dialog should now look like Figure 3 37 119 Chapter 3 Decision Analysis Results Syncopation Software gt Click OK to run the Rainbow Diagram Rainbow Diagram License vs In house downstream Varnable Variable name p Curent value 0 6 From 0 3 To 0 9 Number of steps 13 C Step size 0 05 E valuation method Full tree enumeration Change OK Cancel Help Figure 3 37 Run Rainbow Diagram Dialog for p DPL produces the Rainbow Diagram for p as shown in Figure 3 38 The Rainbow Diagram indicates that while the objective function increases from approximately 102 to 120 no policy changes occur as p varies from 0 3 to 0 9 In this i
446. ve to make Marketing Costs conditioned on Product Strategy by changing the arc type Like Cost per Unit Marketing Costs will be zero if Product Strategy is in the License alternative See Figure 2 24 48 Syncopation Software Chapter 2 Building Decision Analysis Models Reorder General Conditioning Print Page Setup Copy Full Screen wheel Marketing Product In house Costs Strategy es License Figure 2 24 Node Data Input Tree for Marketing Costs gt If you wish bend the influence arc from Product Strategy to Marketing Costs Your model should now look something like Figure 2 25 49 Chapter 2 Building Decision Analysis Models Syncopation Software Product Strategy In house El NPV License lt NPV Figure 2 25 Model with Conditioning from Product Strategy Decision Note in this example you conditioned several nodes on the Product Strategy decision For example you specified that both Cost per Unit and Marketing Costs were zero if the Product Strategy alternative is License Another way to do this would be to link the decision node to the spreadsheet with flag values e g 1 and 2 and putting the logic for marketing costs etc in the spreadsheet e g an Excel formula such as if Product_Strategy 2 0 Marketing_Costs As with any other node type decision nodes can be linked to Excel Now that you have added a decision to your model you will run
447. velop Drug is Yes then Costs is conditioned by Develop Drug See also Dependency Constraint Function An if then else expression that tests a condition and specifies different objective functions for true or false outcomes or prunes the tree eliminating any further branches 461 Appendix D Glossary of DPL and Decision Analysis Terms Syncopation Software Continuous Chance Node A node in an Influence Diagram or decision tree that represents an event whose outcomes are uncertain A continuous chance node does not have a discrete number of states To generate a distribution of outcomes for a continuous chance node samples are continuously drawn from a named distribution The graphical symbols for a continuous chance node are dark green ovals in the Influence Diagram and dark green circles in the Decision Tree Control Branch See Branch Control Controlled Node A node in the Influence Diagram representing an event whose states are set as an action on a branch in a decision tree Controlled nodes do not appear in the decision tree The graphical symbol for a controlled node is a white rectangle in the Influence Diagram A controlled node is different from a controlled branch on a node see Branch Control Conversion DPL can convert spreadsheet models to blocks of DPL code The resulting code can be included in DPL models or edited and run as a stand alone DPL program Cumulative Distribution a graph generated by DPL whose X coo
448. verted spreadsheets run much more quickly than models linked to Excel DPL can evaluate models that combine graphic models and DPL programs so you can continue to use the Model window to build Influence Diagrams and Decision Trees while using a converted spreadsheet e Documenting your model if you need a printed hard copy version of your complete model you can print the Influence Diagram Decision Tree and all the data trees from the Model window Or you can convert your model to a DPL program and print it out The DPL program is usually much more compact e Quality assurance you may find that a DPL program s compact presentation of all the data and formulas in your model makes verifying that everything is complete and correct an easier task DPL can generate programs automatically from Influence Diagrams Decision Trees 15 2 OVERVIEW OF DPL PROGRAM COMPONENTS A typical DPL program is divided into two sections a definition section and a sequence section In the definition section you enter a collection of statements that describe the elements of the decision problem decisions chance events data and how they work together The definition section is analogous to the Influence Diagram In the sequence section you describe the sequence in which decisions are made uncertainties are resolved and values are calculated The sequence section is analogous to the Decision Tree It is not necessary to include both sections in every
449. view of DPL s fundamental decision analysis outputs You may wish to review that chapter before proceeding with this chapter This chapter assumes that you are familiar with the basics of using DPL Professional i e with the material covered in Chapters 2 and 3 of this manual 12 1 POLICY TREE FEATURES You will start by looking at the Format Display features for pruning Policy Trees to make them easier to examine and analyze Select File Open Navigate to the Examples folder underneath where you installed DPL If you used the default location the path is C Program Files x86 Syncopation DPL8 Examples Select Policy Tree da and click Open DPL opens the Workspace as shown in Figure 12 1 341 Chapter 12 Advanced Decision Analysis Results Syncopation Software 9 Jd DPL Policy Tree da Drug Development EN Workspace Manager y y Workspace _ Drug Development Session Log 9 OPL Cecision Programming Language Copyright 2002 2013 by Syncopaten Sotware I P eer r on Seosee teva at r G Paret Key Wiha Kat Market Preine mpal Share Size cocog POSET PRO Some La h Low Low Lo t sy copanon com FDA j j US tobtree 1 666 796 2375 Pa Failure j j PY Market m BE 440 35 LO Oey j 3 j f Quise te US 1 978 233 2 A pper f Yes Nomnal Nominal Nominal Professional versi Yes Modarsta success Pv Launch PV Maket Perpetual es No P
450. w whether you are better off not launching it In this section you ll add a chance node and a decision node to model one such exit option As before switch back to the Asymmetric model Click Model Links Add The Range Names dialog appears Select Key_competitor_outcome Click OK The new node is initially placed to the right Move the new node so that it is above FDA Approval GUGEUY While the node is still selected drop down the Model Node Change to split button on the ribbon see Table 1 2 and choose Discrete Chance from the list The Node Definition dialog appears with the General tab selected 4 Change the outcomes to Failure Disappointment and Breakthrough Click on the Data tab of the Node Definition dialog and enter values of 1 2 3 See Figure 6 24 Leave the default probabilities of 3 4 and 3 203 Chapter 6 Asymmetric Trees Syncopation Software General Conditioning Print Page Setup Copy Il Full Screen Wheel Reorder Probability Failure competitor BEE outcome Disappointment 4 Breakthrough Figure 6 24 Node Definition dialog Data tab Click OK Your model should look like Figure 6 25 204 Syncopation Software Chapter 6 Asymmetric Trees Key Market Market ini competitor Share Size outcome Failure Low Low PV_Market Nominal Nominal Disappointment lt PV_Market High High Breakthrough PV_Market
451. xpand further 252 Syncopation Software Chapter 9 Multidimensional Value Nodes 9 MULTIDIMENSIONAL VALUE NODES Multidimensional value nodes allow you to model series one dimensional arrays and two dimensional arrays in the Influence Diagram As with scalar value nodes these value nodes can contain numeric constants and or formulas Each of these types of multidimensional value nodes will be demonstrated in the sections that follow Regardless of its dimensionality any value node in DPL can be linked to a DPL program or an Excel spreadsheet Value nodes that are linked to Excel spreadsheets can be either import nodes that are calculated by Excel and returned to DPL or export nodes whose data DPL sends to Excel for use in Excel s recalculation This chapter covers three situations a one dimensional array that is an import node a one dimensional series that is an export node and a two dimensional array that is an export node This chapter assumes that you are familiar with building basic DPL models linking DPL models to Excel and naming ranges in Excel If you are new to DPL and or naming ranges in Excel you may wish to review Chapters 2 and 3 of this manual and or your Excel documentation before proceeding with this chapter If you are building cash flow models and wish to analyze the uncertainties associated with your cash flows over time you may wish to work through Chapter 9 1 1 of this chapter and the other sections
452. xpressions formulas Using the Select Variable dialog saves typing reduces errors from misspellings and prevents you from having to remember the names of all the variables in your model DPL variable names are case sensitive Costs is different from costs gt Use the down arrow to move the selection to the entry for License Click the select Variable button M and double click License_Fee in the list Press Enter Your data input tree for Revenue per Unit should now look like Figure 2 19 43 Chapter 2 Building Decision Analysis Models Syncopation Software Node Definition Revenue per Unit A fa S General Conditioning Print Page Setup Copy Full Screen Wheel Reorder Links Revenue Product In house per Unit Strategy License License_Fee Figure 2 19 Data Input Tree for Revenue per Unit Node You have now told DPL that when Product Strategy is in the In house alternative Revenue per Unit is equal to Price whereas when Product Strategy is in the License alternative Revenue per Unit is equal to License Fee gt Click OK to close the Node Definition dialog Your model should now look something like Figure 2 20 Note that DPL has drawn an influence arc from Product Strategy to Revenue per Unit to indicate that Product Strategy conditions Revenue per Unit Also note that the arrowhead of the arc is colored DPL uses a blue arrowhead to indicate that the
453. y the icon f and a Risk Profile dataset by the icon When you request a Risk Profile for the optimal alternative only DPL names the Risk Profile chart and dataset Expected Value DPL encloses items in the Workspace Manager that do not correspond to windows such as the Risk Profile datasets with curly braces i e Expected Value Double click on the item for Risk Profile chart called Expected Value a in the Workspace Manager to activate the window You may also select it and press Enter The Expected Value Risk Profile chart becomes active as shown in Figure 3 7 Note you can also double click on the Risk Profile dataset called Expected Value DPL activates the Risk Profile chart that displays it If you double click on a Risk Profile dataset that is not displayed in a Risk Profile chart DPL creates a chart for the dataset and activates the chart 85 Chapter 3 Decision Analysis Results Syncopation Software Gao CC DPL New Product Strateqy da Expected Value ES Fie Home Model View Date Senet format Help let 4 Y lines Markers show p Modify t Weignt bir Labe t Workspace Manager 9 lt 8 New Product Strategy da x License vs In house a liceme vs in house prob Expected Value J Gxpected Value ig Policy Tree Tornado Tornado3 5 9 c be a w gt Session Log 9 5 OPL Decision Programming Language Copyright 2003 2013 by Syncopation Sa 3 incorporating sof
454. y Data You can change any of the probabilities in your model and play endpoints instead of re running the entire model You will change the probability distribution for Actual Reserves to be more optimistic Edit the probability data for Actual Reserves so that it looks like Figure 11 9 Note that only the probabilities not the values are changing 317 Chapter 11 Using the Endpoint Database Syncopation Software Node Definition Actual Reserves General Conditioning Print Page Setup Copy Full Screen Wheel Reorder Multiple Experts Q AQ Distribution type discrete v Links Probability 6 vis Value 5000 vis Actual Reserves Figure 11 9 Revised Probabilities for Actual Reserves Node gt Click OK Note that although Full Tree Enumeration from Endpoints or playing endpoints was the last evaluation method used DPL does update the Run Decision Analysis split button Therefore each time you wish to run a Full Tree Enumeration from Endpoints you must drop down the Home Run Decision Analysis split button and select it from the list or press the keyboard shortcut AIt F10 Run a Full Tree Enumeration from Endpoints gt Click Ok to the warning The expected value has increased substantially to 46 8 gt Switch back to the Model Window and select Undo from the Quick Access menu 318 Syncopation Software 11 2 3 Chapter 11 Using the Endpoint Databas
455. y location r j i Show ranges already linked 7 Show ranges with formulas Create arrays based on range ize Prefix node name with sheet name Figure 5 27 Range Names dialog Click OK DPL initially adds the new nodes to the right gt Move them so they are positioned as in Figure 5 28 171 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software Figure 5 28 Influence Diagram with Spin off and Loss of Synergy added Next you need to change Spin off into a decision node Select Spin off and drop down the Model Node Change to split button and choose Decision from the list DPL displays the Node Definition dialog The default decision alternatives Yes and No are fine but you need to edit the data In the spreadsheet a value of 0 for Spin off means that no action is taken and a value of 1 means that Enlightify is sold gt Click on the Data tab in the dialog Press the down arrow to navigate to the No branch gt Type 0 Figure 5 29 gt Click OK 172 Syncopation Software Chapter 5 Building and Analyzing Monte Carlo Simulation Models General Conditioning Print Page Setup Copy Full Screen Wheel Reorder 1 Figure 5 29 Node Definition Data for Spin off gt Select Model Influence Arc From Formulas so that DPL will add arcs for the new nodes Your model should l
456. y1 xls Compatibility Mode Microsoft Excel Home Insert Page Layout Formulas Data Review View DPL B amp Arial 7110 General 7 E Conditional Formatting g Insert AP Fy oe oe ae Oe ee ee ty W 2 Format as Table 7 g Delete g Z l de 7 H 2 Ae Xor ai a E Cell Styles 7 Bj Format 2 ae obec Clipboard M Font F Alignment amp Number F Styles Cells Editing a i Wt Enlightify Consulting LLC Assumptions Basics Discount Rate Interest Rate on Working Capital Valuation Date Results Total Value m Uncertainties and Decisions Trendline Revenue Growth Rate Bad Debt Allowance of Revenue Share of Revenue at Risk due to New Regulations Share of at Risk Revenue Lost Gross Margin Working Capital days Other Cost Growth Rate RMC Lawsuit Settlement m tions DCF 2 Figure 5 2 Excel Cash Flow Model for Enlightify Assumptions Sheet This simple cash flow model calculates the value of Enlightify Consulting LLC a troubled business unit whose financial performance is of concern The Excel model is built on two sheets one for key assumptions and value drivers and another for the cash flow calculations 143 Chapter 5 Building and Analyzing Monte Carlo Simulation Models Syncopation Software In Figure 5 2 you can see that eight of the value drivers have already been identified as sources of uncertainty the cells are shaded in light green Each of these cells

Download Pdf Manuals

Related Search

Related Contents

電気メス用付属品（未滅菌）  User Manual  Dynnox F  Palm 700p User Guide  （第29期）第2四半期決算説明会資料 （PDF：400 KB）  メーヨー・ヘガール持針器  Implementation of CAN Bus Protocol  Manuel de l`utilisateur  EZ-Ject®, Inc. EZ-Ject®, Inc. - EZ-Ject Herbicide System and Soil  

Copyright © All rights reserved. Failed to retrieve file