Improve Analytic Results with Decision Modeling
3 likes3,796 views
The document discusses strategies for improving analytic results through decision modeling, emphasizing the importance of understanding decision-making processes and requirements. It outlines a four-step approach for identifying, describing, and specifying decision requirements to enhance analytics deployment and effectiveness. Key elements include collaboration across organizations, clarity in decision-making, and the need for detailed decision documentation to drive better business outcomes.
Improve Analytic Results with Decision Modeling
- 1. IMPROVE ANALYTIC RESULTS USING DECISION MODELING James Taylor Decision Management Solutions
- 2. • CEO of Decision Management Solutions • I work with clients to improve their business by applying analytic technology to automate & improve decisions • I have spent the last 11 years championing Decision Management and developing Decision Management Systems ©2013 Decision Management Solutions 2 YOUR PRESENTER – JAMES TAYLOR
- 3. ©2013 Decision Management Solutions IMPROVING ANALYTIC RESULTS 3
- 4. 4 LIMITS ON ANALYTIC SUCCESS ©2013 Decision Management Solutions Poor Understanding Cottage Industry Practices Business Poor Collaboration Ineffective Deployment 4
- 5. • Decision Makers ask for data or an analytic • But this does not help them make the decision • So they ask for more or different data • But that doesn’t help either • So they ask for a different view of the data • … • Until they get fed up and do something in Excel! ©2013 Decision Management Solutions 5 THE TELEPHONE GAME
- 6. MULTIPLE PERSPECTIVES REQUIRED ©2013 Decision Management Solutions 6 Business
- 7. ANALYTICS MUST BE DEPLOYED ©2013 Decision Management Solutions 7 ? ?
- 8. INDUSTRIAL SCALE ANALYTICS ©2013 Decision Management Solutions 8
- 9. ©2013 Decision Management Solutions DECISION MODELING 9
- 10. “Knowing is not enough; we must act.” Johann Wolfgang von Goethe ©2013 Decision Management Solutions 10 WHY MODEL DECISIONS?
- 11. • Identifying decisions in workflows does not define decision making • Workflows show • When decisions are made • Who makes decisions • What information they use • What the consequences are • Workflow doesn’t show HOW ©2013 Decision Management Solutions 11 DECISION AND WORKFLOW THINKING
- 12. • Adding Decision Modeling explains HOW a decision is made • Explicitly shows • Information used • Questions answered • Policies, regulations used • Role of analytics ©2013 Decision Management Solutions 12 DECISION AND WORKFLOW THINKING
- 13. • Identify Decisions • Describe Decisions • Specify Decision Requirements • Complete Requirements ©2013 Decision Management Solutions 13 4 STEPS TO SUCCESS
- 14. • Identify Decisions • Describe Decisions • Specify Decision Requirements • Complete Requirements ©2013 Decision Management Solutions 14 4 STEPS TO SUCCESS
- 15. TWO SCENARIOS Business Area Focus Decision Model finds opportunities for improvement with analytics Specific Analytic Decision Model ensures clarity, traceability and reuse ©2013 Decision Management Solutions 15
- 16. Business Processes DISCOVERING DECISIONS ©2013 Decision Management Solutions 16 Business Intelligence Brainstorm KPIs Micro and hidden Decisions
- 17. • Knowing what a business user will do with information points at decisions Ask “So what?” • Potential options available for a choice identify possible decisions Identify Actions • Places where significant branching points to one path or another Business Processes • Analytics are often powerful when used to differentiate treatmentAsk “Why?” ©2013 Decision Management Solutions 17 FIND DECISIONS
- 18. TARGET DECISION-MAKING ON KPIS ©2013 Decision Management Solutions 18 Choices, decisions affect KPIs KPIs motivate behavior Strategy defines KPIs
- 19. • Identify Decisions • Describe Decisions • Specify Decision Requirements • Complete Requirements ©2013 Decision Management Solutions 19 4 STEPS TO SUCCESS
- 20. • Name • Description • A question • A defined set of allowable answers • Any other results to be returned with the answer • Key facts like volume, complexity, latency ©2013 Decision Management Solutions 20 INFORMATION ABOUT A DECISION Repeatability One-off Continuous Measurability Poor Specific Time to outcome Immediate Delayed Volume Low High Value Range Narrow Gap Wide Gap Complexity Low High
- 21. • Be specific with questions • Subject • Timing • Scope or limitations ©2013 Decision Management Solutions 21 QUESTIONS Decision Good question Bad question Customer Retention What retention offer can we make this customer when they call to cancel their service? How can we retain this customer? Supplier Selection Which of our approved suppliers should be selected for this specific parts order? What supplier should we use? Preventative Action What is the prioritized list of preventative actions for this quality team on this line today? What preventative action should the quality team take?
- 22. ALLOWED ANSWERS • Supporting Information • Messages • Warnings • Notes • Explanations • Single Answer • List of Options • Structure with parts ©2013 Decision Management Solutions 22 Type Description Notes Yes/No Yes or No Or True/False, 1/0 etc Number A numeric value Often constrained to a value in a specific range Specific Value One of the values specified in a list For example Accept / Reject / Refer Database value A value stored in a database Specify how to get the list of options - products, pieces of content etc. Other Generally a string or block of text Such as a custom script or personalized email body Structure A set of values each of which is of one the allowed types Some decisions involve the rolled-up output of their component decisions.
- 23. DECISION TO KPI MAPPING ©2013 Decision Management Solutions 23 KPIs impacted by Decisions Customer Retention Customer Profitabilit y … Select Marketing Offer X Determine Discount X X Determine Shipping Mechanism X … X Table 1: Impact of Project on Objectives Linked to These Decisions Objectives to be impacted Impact Customer Retention Rate Improve as much as possible Customer Profitability Keep constant within 2% of baseline
- 24. • You need to know who has to approve your analytics • You need to know who has to believe your analytics • Mismatches between organizations can be very problematic ©2013 Decision Management Solutions 24 ORGANIZATIONAL CONTEXT MATTERS
- 25. • Identify the Organizations that specify how the decision should be made • They care about the approach Own • Identify the Organizations that make these decisions day to day • They care about execution Make • Identify other impacted Organizations e.g. those whose KPIs are impacted • They care about results Impacted ©2013 Decision Management Solutions 25 THREE ORGANIZATIONAL ROLES
- 26. BUILD CONFIDENCE WITH CONTEXT • Business Processes • Business Events • Existing Systems • What processes does this business execute? • Which business processes will need decisions made? • What key business events trigger business behavior? • When will decisions be required and in what context • What are the core systems for this business area? • How will you have to deploy decision making? ©2013 Decision Management Solutions 26
- 27. • Identify Decisions • Describe Decisions • Specify Decision Requirements • Complete Requirements ©2013 Decision Management Solutions 27 4 STEPS TO SUCCESS
- 28. Information • What is needed? • Where does it come from? Know-how • How to make it • How to improve it Precision • Exactly how? • Specificity without technical details Context • Application • Organization • Business Goals ©2013 Decision Management Solutions 28 DESCRIBING A DECISION
- 29. • Model Information Requirements • Information about the case • Transaction details • Reference information • For example • “Determine Parts Availability” requires BOM and Inventory information • “Validate Tax Return” requires Return and Citizen information • “Refer claim for fraud” requires Claim and Provider information ©2013 Decision Management Solutions 29 INFORMATION REQUIREMENTS
- 30. • Model Authority Requirements • Policies or regulations • Expertise • Analytic Insight • For example • “Reorder parts” requires knowledge of supplier capabilities and risks of shortage • “Refer claim for fraud” requires the likelihood of fraud and state regulations ©2013 Decision Management Solutions 30 AUTHORITY REQUIREMENTS
- 31. MORE THAN ANALYTICS © Decision Management Solutions, 2013 31 Decision
- 32. HIGH LEVEL REQUIREMENTS ©2013 Decision Management Solutions 32 Decision KnowledgeInformation
- 33. • What is required to make decision? • Information • Guidelines, policy • Expertise • Regulations • Predictive Analytic Models • Data Mining Results • The results of other decisions ©2013 Decision Management Solutions 33 DECOMPOSE THE DECISIONING
- 34. • Repeatable decisions • Specifying a kind of decision not a specific decision • Typically arranged in a tree but actually a network • Lower-level decisions often reused by higher-level ones • Core decisions layered channel-specific decisions • Directional links between nodes • If A is linked to B then B requires A • Cyclic links are not allowed ©2013 Decision Management Solutions 34 DECISION REQUIREMENTS DIAGRAM
- 35. ©2013 Decision Management Solutions 35 ADD MORE DETAIL
- 36. ©2013 Decision Management Solutions 36 DETAILED DECISION DIAGRAMS Information Knowledge Decision
- 37. LINK INFORMATION AND INSIGHT ©2013 Decision Management Solutions 37 DecisionKnowledgeInformation
- 38. • Identify Decisions • Describe Decisions • Specify Decision Requirements • Complete Requirements ©2013 Decision Management Solutions 38 4 STEPS TO SUCCESS
- 39. • Introduction • Objectives/KPIs Impacted • Analytics Required • Analytic Approach • Information Sources Used • Decisions Impacted • Application context • Systems, Business Processes, Events • Organizations involved • Owners, makes and impacted • Other Project details ©2013 Decision Management Solutions 39 ANALYTIC PROJECT DOCUMENTATION Table 1: Impact of Project on Objectives Linked to These Decisions Objectives to be impacted Impact Customer Retention Rate Improve as much as possible Customer Profitability Keep constant within 2% of baseline
- 40. ©2013 Decision Management Solutions DECISION MODELING AND ANALYTICS 40
- 41. BUSINESS INTELLIGENCE AND ANALYTICS ©2013 Decision Management Solutions Adapted from Competing on Analytics, Davenport and Harris, 2007 41
- 42. BUSINESS INTELLIGENCE AND ANALYTICS ©2013 Decision Management Solutions Adapted from Competing on Analytics, Davenport and Harris, 2007 42 What decision will you make when you get this report? What are you doing when you consult it?
- 43. BUSINESS INTELLIGENCE AND ANALYTICS ©2013 Decision Management Solutions Adapted from Competing on Analytics, Davenport and Harris, 2007 43 Who has to make a decision? How much of the decision can the system make?
- 44. BUSINESS INTELLIGENCE AND ANALYTICS ©2013 Decision Management Solutions Adapted from Competing on Analytics, Davenport and Harris, 2007 44 If we knew this what could we do differently? How will we use these analytics to improve results?
- 45. ©2013 Decision Management Solutions WRAP-UP AND QUESTIONS 45
- 46. Decision Models clarify business understanding Decision Models are shared between business, IT and analytic teams How the analytic will be deployed and used is clear Decision Models scale up and formalize the definition of business understanding ©2013 Decision Management Solutions 46 ANALYTIC SUCCESS WITH DECISION MODELS
- 47. Social, collaborative, cloud-based Decision Management modeling software that puts decisions at the heart of your business architecture, building better requirements and driving successful analytics implementations ©2013 Decision Management Solutions 47 DECISIONSFIRST MODELER
- 48. • To contact James • james@decisionmanagementsolutions.com • +1 (650) 400 3029 • JTonEDM.com • To learn more about Decision Management and DecisionsFirst Modeler • decisionmanagementsolutions.com • decisionmanagementsolutions.com/decisionsfirst-modeler • To learn more about IIA • iianalytics.com ©2013 Decision Management Solutions 48 MORE INFORMATION
- 49. research@iianalytics.com 49©2013 Decision Management Solutions
Editor's Notes
- #2: In this webinar, IIA Faculty Member James Taylor, CEO of Decision Management Solutions, will show how to improve analytic results with decision modeling. Decision modeling focuses analytic efforts, clarifies the business goals of analytic projects, and improves collaboration between analytic, business and IT organizations. James will introduce decision modeling, show how it can be used in a wide range of analytic projects and share experiences from using decision modeling in various industries.
- #6: Lots of iteration because the providers of data/reporting/analysis are remote from the decision makers and have no way to develop a shared understanding of the decision making being assistedAsk for data which they get but doesn’t get the answer so asks for more data etc, telephone game of sorts
- #7: Business, IT and Analytics team can collaborate effectively
- #9: Adopt an industrial mindsetEach model is hand-crafted - Expertise is applied in an automated contextScripts and programming are primary - Graphical analytic tools are primaryModels are one-time efforts - Models are continuously refreshed and updatedProjects are done when the model is done - Projects are done when the business is changed
- #19: A second approach is to look at the Key Performance Indicators (KPIs) and metrics that the business has for the area under consideration. There might be a specific KPI or metric that is being targeted by the project. Even if there is however it is worth identifying the other metrics and KPIs in the business area as any decision is likely to impact several and any KPI or metric can be a good source for decisions.Any KPI or metric is valuable only if it helps motivate suitable behavior and that implies that someone can change the value of that KPI or metric. By investigating KPIs and metrics, and finding out when and where people make choices that move KPIs or metrics up or down, a project team can identify decisions. Each opportunity for choice-making, for selecting an action from a possible set of actions, is a decision.To begin the team can simply ask what decisions make a difference to a KPI or metric but they may find they have to ask business experts to walk through their day or week keeping the KPI in mind to find all of them as people may not think of the choices they make as explicit decisions.