Credit Risk Modeling and Decisioning Conference Larry Rosenberger May 30, Strategy Science: Breakthrough in Data-Guided Decision-Making

Transcription

1 Credit Risk Modeling and Decisioning Conference Larry Rosenberger May 30, 2002 Strategy Science: Breakthrough in Data-Guided Decision-Making

2 T O P I C S Basics of Strategy Science Recent Lessons from the Field Speculations on the Future Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. Credit Modeling and Decisioning Conference 2

3 T O P I C S Basics of Strategy Science What s the breakthrough? What is Strategy Science? Extending existing notions Recent Lessons from the Field Speculations on the Future

4 Current Industry Practice: Decision strategy design driven by judgment Information External Data ANALYTICS Predictions D A T A D A T A DATA MGMT. Predictive Analytics Exploratory Analysis / Data Mining Judgmental Strategy Design AUTOMATED STRATEGY EXECUTION Internal Data Insights Reactions Actions Prospects & Customers Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 4

5 The Breakthrough: Adding science to strategy design Information Optimal Decision Strategies External Data ANALYTICS Predictions D A T A D A T A Internal Data DATA MGMT. Predictive Analytics Exploratory Analysis / Data Mining Insights Strategy Science Decision Modeling Strategy Optimization Strategy Refinement AUTOMATED STRATEGY EXECUTION Reactions Actions Prospects & Customers Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 5

6 Strategy Science 3 Step Methodology Replaces guesswork in strategy design with science Decision Modeling Strategy Optimization Build a graphical model for one or more decisions Establish mathematical relationships Solve for profit-improvement improvement strategies - subject to constraints on key metrics Strategy Refinement Risk R evenue Incom e Expert D ecision Product Low Low Low Low Low Low Medium Med Med High High High High Low Low High High Med Med Low Low High High High High D D D A B /C A /C A Refine strategies for interpretability, robustness and ease of deployment Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 6

7 Simplified Decision Model for On-going Credit Line Decision models drive Strategy Science Current Balance $ Rev Credit Bureau Behavior Good / Bad Credit Line Change Profit $ Loss Current Limit Modeling your customers reactions to your actions is key! Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 7

8 Optimization and refinement produce a better decision strategy Risk Low Medium High Utilization Low Med High Low Med High Balance Low High Low High Expert Decision $5,000 $8,000 Optimal Decision $3,000 $5,000 $12,000 Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 8

9 Strategy Science Extending existing notions Credit Scoring better PREDICTIONS Strategy Science better DECISIONS Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 9

10 Predicting vs. Decisioning Prediction focus Profit Profit Response Response Revenue Revenue Risk Risk D A T A D A T A D A T A D A T A D A T A Judgmental Decision Strategies Expert Judgment Prospects & Customers Actions Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 10

11 Predicting vs. Decisioning Decision focus Profit Profit Response Response Revenue Revenue Risk Risk D A T A D A T A D A T A D A T A D A T A Strategy Science = Decision Modeling + Strategy Optimization + Strategy Refinement Optimal Decision Strategies Prospects & Customers Actions Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 11

12 Strategy Science Extending existing notions Credit Scoring better PREDICTIONS Strategy Science better DECISIONS Risk Management better loss control Strategy Science better trade-offs Strategy Science better understanding of trade-offs Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 12

13 Trading-off volume versus losses Baseline operating point Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 13

14 Profit implications of the volume vs. loss trade-off Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 14

15 Creating the Efficient Frontier Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 15

16 Exploring Trade-offs Among Key Business Metrics Baseline Profit / account $73 Profit / account $95 (+30%) Profit / account $78 (+7%) Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 16

17 Basics of Strategy Science T O P I C S Recent Lessons from the Field Where are we? Scar tissue Speculations on the Future

18 Where Are We? The 3 rd Decision Revolution today Profit Improvement Predictive Modeling 20-30% Adaptive Control 20-30% Strategy Science 20-30% Time Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 18

19 Where Are We? FAST adoption! Within 1 year after launching Strategy Science, 2 countries: U.S., U.K. 3 credit products: credit card, direct & indirect loans 8 decision areas in the customer life-cycle 26 projects in progress/completed for 16 clients Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 19

20 Applied Strategy Science Decision areas across the customer life-cycle Strategy Science for Financial Services Customer Acquisition Decision Models Customer Management Decision Models Supporting Analytics Bankcard Pre-Screen Intro APR Go-to APR Intro period Rewards offer Affinity offer Credit line assignment Installment loan terms and pricing Indirect auto loan terms and pricing Bankcard Credit line management Authorizations (early stage collections) Collections Retention and Loyalty programs Cross-sell Capabilities Data Spiders ClusterBots Utility Functions Methodologies Pre-Market Offer Testing Learning Strategies Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 20

21 Where Are We? Early results to date On target to achieve 20-30% profit improvement in this 3rd Decision Revolution Early field tests outperforming benchmark strategies Some early adopters have gone beyond their first application of Strategy Science Fast followers are paying serious attention It s too early for proven only players Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 21

22 Scar Tissue Hard stuff really hard stuff Success requires large doses of: Mathematical modeling art Data analysis savvy Domain expertise injection Success forces numerous innovations Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 22

23 Hard Stuff Really Hard Stuff Mathematical modeling art Insightfully model customer reactions to lender actions Insightfully choose key business metrics for trade-off analysis of conflicting business objectives Insightfully represent business constraints for iterative optimization Balance simplicity versus complexity Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 23

24 Hard Stuff Really Hard Stuff Data analysis savvy Estimate quantitative relationships based on both data and expert judgment to account for: Holes in the data Confounding effects of multiple decisions Exogenous variables and influences Estimate uncertainty of estimates Balance simplicity versus complexity Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 24

25 Hard Stuff Really Hard Stuff Domain expertise injection Create tools and methodologies to inject business judgment flexibly and at multiple points Engineer resulting strategies for: Transparency/Interpretability Robustness Ease of deployment Balance simplicity versus complexity Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 25

26 Hard Stuff Really Hard Stuff Forced innovations Consumer preferences Stated preferences/intentions Observed preferences/revealed intentions Action-based prediction Existing tools but new methodologies Learning strategies new role for experimental design Efficient horse racing versus horse breeding Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 26

27 Basics of Strategy Science Recent Lessons from the Field T O P I C S Speculations on the Future Technical directions Impact on decision-making

28 Where Is Strategy Science Going? Technical directions Inject feed-forward control in the face of dynamics! Economic stress testing Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 28

29 Feed-Forward Forward Control Simulating different economic conditions Current Balance Simulate expected Simulate expected changes in the changes in the model inputs model inputs Alter Alter constraints constraints $ Rev Credit Bureau Behavior Good / Bad Credit Limit Change Profit Alter output Alter output Current Limit Alter the relationship Alter the relationship between model between model elements elements $ Loss Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 29

30 Where Is Strategy Science Going? Technical directions Inject feed-forward control in the face of dynamics! Economic stress testing Model many decisions and integrate tightly-coupled ones Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 30

31 Model Many Decisions And integrate the tightly coupled ones Potential Customer Rejected Prescreen Decision Solicit Decision No Response Prescreen Accepted Responder No Response Contact Decision Responder New Customers Prospects Approve, Product Offer Decision Accept Product Don t Accept No Activation Product Offer and Activation Enhancement Activation Enhancement No Activation Account Management -Credit Lines -Pricing -Payment Policies Customer Service Inactive Customers Reactive Inactive Customers Decision Retention, Usage Decision User Payer Active Customers Current Customers Delinquent Customers Payment Management Decision Portfolio Acquisition Portfolio Sale Attriter Closed Account Defaulted Customers Former Customers Collection Data Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 31

32 Where Is Strategy Science Going? Technical directions Inject feed-forward control in the face of dynamics! Economic stress testing Model many decisions and integrate tightly-coupled ones Design and build multi-tier control systems Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 32

33 Multi-Tier Decision Control Systems Function line of business enterprise Card Portfolio ENTERPRISE Profit Portfolio Portfolio Objectives Objectives Product Product mix mix Profit Profit contribution Loss Loss reserves reserves Acquisition Acquisition growth growth Trade-off Levers Customer mix Market growth Capital allocation Capital reserves Market Share EPS Pre-Screen Initial Credit Line Credit Line Collections Business Objectives Acct acquisition cost Response Portfolio growth Business Objectives Activation Contingent liability Retention Balance transfer Business Objectives Retention Receivables Contingent liability Losses Business Objectives Dollars recovered Collection costs Third party mgt Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 33

34 Where Is Strategy Science Going? Changing decisions and decision-making processes Quantifying trade-offs Operating over the cliff! But how far? Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 34

35 Quantifying Trade-offs Simplified indirect auto loan pricing Risk score Accept/ Accept/ Reject Reject Offer Offer Custom Risk Take/ No Take Profit Market Data Deal Data Response Good/ Bad Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 35

36 Efficient Frontier The staircase to credit loss hell high Low cutoffs E [ Volume ] High cutoffs low high Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 36

37 A Closer Look at an Efficient Frontier Profit vs. Market Share ( Take Rate ) Totalprofit vs.take ra te (for 1 m illion prospects) I Totalprofit (in m illions) Historical o p e ra tionalpoint Totalprofit II III IV 0% 10% 20% 30% 40% 50% 60% Take rate Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 37

38 Where Is Strategy Science Going? Changing decisions and decision-making processes Quantifying trade-offs Operating over the cliff! How far? Changing dialogs & roles of key players data-assisted BOGSAT technology Business objectives & trade-offs Recent changes in the marketplace Future uncertainties Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 38

39 Changing Dialogs and Roles of Key Players Data-assisted assisted BOGSAT decision process Imagine being able to explore the impact of policy changes, the sensitivities of your business constraints, and the forces of economic and competitive change on any decision strategy before it s implemented. Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 39

40 Conclusions Strategy Science is HUGE! Early adopting continues, while fast following is starting-up It s really hard stuff! Copyright 2002 Fair, Isaac and Company, Inc. All rights reserved. 40

41 Strategy Science: Breakthrough in Data-Guided Decision-Making Questions & Answers