Predictive Analytics: Achieving Greater Decision Accuracy, Better Risk Segmentation, and Greater Profitability Lamont D. Boyd, CPCU, AIM Insurance Market Director FICO Scoring Solutions LamontBoyd@FICO.com October 2012 1 This presentation is provided for the recipient only and cannot be reproduced or shared without Fair Isaac Corporation's express consent.
Agenda» Overview/Research Findings» 128 patents predictive analytics and decision management» Insurance Scores» Credit-Based Insurance Scores» Property Risk Scores» Safe Driving Scores» Custom Predictive Analytics/Custom Modeling» Risk Segmentation for Marketing, Underwriting, Product Development, Pricing, Claims..» Next Steps» Q&A 2 Confidential.
Predictive Analytics: Achieving Greater Decision Accuracy, Greater Profitability Towers Watson Survey (4Q/2011)» Percentage of insurers currently using or planning to use predictive modeling for underwriting and/or pricing» 88% - Personal Auto» 83% - Homeowners» 73% - Commercial Property/CMP/BOP» 72% - Workers Comp» 63% - Commercial Auto» 52% - General Liability» Why the growth in predictive modeling?» 85% - Rate accuracy» 69% - Loss ratio improvement» 69% - Profitability 3
Predictive Analytics: Achieving Greater Decision Accuracy, Greater Profitability SMA (Strategy Meets Action) Research (2012)» 75% of P&C insurers will increase spending on analytics over the next three years» 60% currently using and 25% implementing analytics in underwriting to better understand and manage risks» 50% currently using and 22% implementing analytics for distribution management/agent performance» 40% using and 15% implementing analytics for claim fraud detection and prevention» Personal lines carriers/units more aggressively using and pursuing predictive analytics than commercial lines counterparts» Noted barriers to capitalizing on predictive analytics» Lack of strategy» Data accessibility/legacy core systems» Lack of priority or funding 4
Agenda Insurance Scores» Insurance scores: CBIS and property risk scores» Custom models: GAM versus GLM» Premium leakage and pre-claims fraudulent detection at underwriting detection activity at underwriting 5 2010 Fair Isaac Corporation. 2010 Fair Isaac Corporation. Confidential. Confidential.
Credit-Based Insurance Scores» Key industry success!» FICO introduced Credit-Based Insurance Scores (CBIS) to the industry in 1993» CBIS scores currently used by ~95% of all auto and home insurers» FICO CBIS scores are monitored for effectiveness re risk segmentation and amended as required by regulatory/legislative mandates» Current generation FICO CBIS scores» Experian/Fair Isaac Insurance Score 2.0 (via Experian/LexisNexis)» InScore 3.0 (via Equifax)» Fair Isaac Insurance Risk Score 2.0 (via TransUnion)» Canadian Property Loss Score (via Equifax Canada) 6
Update: FICO CBIS in Current Economic Environment Key Question: Have FICO CBIS risk segmentation performance/score distributions changed over time?» Ongoing FICO and client analysis reveals continuing high levels of performance» FICO CBIS score distributions remain essentially the same» Some consumers are seeing lowering scores» Economic climate-driven factors» Majority of consumers are seeing slightly rising scores» More cautious credit management practices» No late payments» Opening fewer accounts» Lowering utilization levels 7
CBIS Usage» CBIS usage provides consistent link in all areas that require strategic decisions Decision Engine Command Center Credit-Based Insurance Score Customer Acquisition New Business Underwriting Tier Placement Renewal Underwriting External Info Purchase Pricing CRM Cross-Selling Billing / Reinstatement Fraud Detection Underwriter Management Field Office Management Production Source Management 8
Acquisition/Marketing Client Benefits» New Customer Acquisition/Marketing Strategy» Focused application» Cost-effective acquisition strategies» Targeting most likely responders» Response modeling» Seeking greatest profit potential» Focused/limited utilization currently» Few regulatory restrictions 9
Underwriting and Pricing Client Benefits» New Business Underwriting Strategy» Earliest CBIS application» Consistent and objective new business risk segmentation decisions» Quickly accepting applicants with greatest profit potential» Dedicating resources to appropriate risks» Minimizing exposure to riskier applicants» Regulatory considerations» Pricing Strategy» Currently focused CBIS application» Slotting applicants by price/tier based on multiple risk characteristics» Gaining premiums relative to presented exposures» Consistent and objective pricing decisions» Regulatory considerations 10
New Business Underwriting Strategy 140 Decline Referral Automated Approval Loss Ratio 130 120 110 100 90 80 70 60 50 40 Average 0 1 2 3 4 5 6 7 8 9 10 Population by deciles 11
Renewal Underwriting Client Benefits» Renewal Underwriting Strategy» Secondary CBIS risk segmentation application» Updating CBIS scores at renewal where allowed» Identifying renewal policyholders for swift processing» Highlighting renewal policyholders requiring greater attention» Renewal tier-placement and pricing relative to exposure» Consistent and objective renewal underwriting decisions» Regulatory considerations 12
CBIS Enhanced Uses Client Benefits» CBIS enhances decision-making and profitability for the enterprise» Book management» Managing books of business at varying level for profitability» Underwriting management» Managing underwriter responsibilities to match risk complexity and varying levels of production and profitability goals» Distribution source management» Managing at agency or other production source levels to assure a focus on greater profitability through better managed and strengthened relationships 13
Regulatory Environment Forecast» Regulatory/legislative questions will continue» Individual state actions» NAIC committees/task forces» FTC Study Auto (2007) and Home (2012/2013)» Education is the key» Education to help consumers understand and change habits to influence credit-based insurance scores is available at www.insurancescores.fico.com» Education to help consumers understand and improve their credit habits to influence the FICO scores lenders use is available at www.myfico.com» Under the 2003 Fair and Accurate Credit Transactions Act (FACT Act), consumers can access each credit report annually via www.annualcreditreport.com 14
Property Risk Scores» Property PredictR Insurance Scores launched in 2007 to help insurers harness predictive value from property inspections performed by FICO s partner Millennium Information Services» Snapshot of risk based on property inspection data» Rank-orders properties by loss ratio relativity» Precise and objective measurement of a property risk» Property PredictR provides strong additive value to CBIS or where regulatory constraints make credit difficult to use» Property PredictR scores are used as a key risk predictor in conjunction with other risk variables 15
Property PredictR Score Retro Analysis Loss Ratio Relativity vs Score Results 1.6 1.53 1.4 1.26 Loss Ratio Relativity 1.2 1.0 0.8 0.6 0.4 1.10 0.92 0.86 0.76 0.64 0.2 0.0 Below 180 11.7% 180-201 9.1% 202-222 19.3% 223-244 17.6% 245-272 20.7% Score by Percent Premium 273-293 11.1% 294 or more 10.5% 16
CBIS with Property PredictR Loss Ratio Relativity Matrix Property PredictR Score SCORE RANGE Credit-Based Insurance Score SCORE RANGE Lowest Scoring 20% Medium Low 20% Medium 20% Medium High 20% Highest Scoring 20% Missing Credit Score Total Lowest 20% 1.83 1.65 1.52 1.85 1.41 Medium Medium Low Medium High 20% 20% 20% Consistently Poor Risks 0.97 0.73 0.65 0.63 0.86 0.68 0.73 0.59 0.58 Consistently Good Risks 1.83 1.11 1.18 1.10 0.93 0.87 Highest 20% 0.66 0.66 0.47 0.71 0.70 Total 1.48 0.97 0.89 0.77 0.56 1.37 1.00 Poor Risks Medium Risks Good Risks Note - Statistic listed in cell is the loss ratio relativity for that cell 17
Safe Driving Scores UBI/Telematics» FICO Safe Driving Scores available in partnership with DriveFactor, a leading telematics provider» Continuous view of risk based on driving behavior» More closely align pricing to loss potential» Accurate measurement of an automobile risk» Consumer-facing views to gradually improve driving habits» UBI/Telematics is the next generation of predictive analytics 18
Agenda» Insurance scores: CBIS and property risk scores Custom Predictive Analytics/ Custom Modeling» Custom models: GAM versus GLM» Premium leakage and pre-claims fraudulent detection at underwriting detection activity at underwriting 19 2010 Fair Isaac Corporation. 2010 Fair Isaac Corporation. Confidential. Confidential.
Insurance Analytic Solutions Insurance Industry Property & Casualty Life Health Rules Mgmt, UW Profit, Claims P&C Personal P&C Commercial Private Payors Government Auto Bureau Scores Risk Scores UW Decisioning Rules Mgmt. Homeowners Workers Comp Property & Liability Rules Mgmt. Fraud Detection Federal Medicare State Medicaid Fraud Detection Bureau Scores Loss Reserving Risk Scores Subrogation Rules Mgmt. UW Decisioning Risk Scores Rules Mgmt. Fraud Detection Fraud Detection Rules Mgmt. 20
Model Development Methodology Attention to detail + experience = superior results Project Design Start Up documentation Customer & FICO Project Team Meeting Project Specifications Analytic Review Analytic Documentation Performance Definitions Performance Def Options Final Performance Definitions Review & Selection Timing Analysis Segmentation Analysis Sampling Plan Sample Selection Pull & Audit Data Analysis Development & Implementation Characteristic Selection Generate Complex Characteristics Model Development & Testing Delivery Meeting Implementation/Tracking Plan 21
Analytics Spectrum Different decisions lead to different methodologies Profiling & Segmentation Predictive Models or Scores Multi-Dimensional Trade-Off Assessment Decision Optimization HIGH X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X Benefit Benefit Establishes broad Establishes segments broad based segments on customer based on profile customer data profile data Rank orders prospects Rank orders on a prospects single dimension on a focused dimension Creates micro Creates segments micro by matrixing segments 2 by or 3 matrixing predictive 2 models or 3 predictive models Brings all predictive Brings analytics all into predictive a single analytics decision into framework a single decision framework Assigns the optimal Assigns action the for optimal each action prospect/account for each prospect/account given specific business given specific constraints business constraints 22
Custom Analytics/Modeling A model that will improve decisions should be» Flexible» Can capture complex data relationships» Palatable» Interpretable» Satisfies operational constraints» Legal concerns» Implementation friendly» Able to be computed quickly and seamlessly deployed» Robust into the future» Limited or graceful performance degradation over time 23
Project design Understanding business challenges» Data Mining - Try multiple algorithms and see which one best matches the business challenge Data Mining Access Data Prepare Modelling Datasets Visualise & Explore Build Standard Models Build Scorecard Models Forecast Predictive Performance Deploy Analytics in Decision Decision» Objectives - Understand the business problem and design an algorithm to predict with precision yet address practical issues 24
Model Development Initial data analysis Fine Binning Coarse Binning» What are we trying to predict?» Metrics» How do we know the model is working?» What data sources are available?» What are costs vs. benefits?» Common data issues» Incomplete and/or inaccurate data» Outlier anomalies can skew statistics» Highly correlated items blur the incremental value contributed by each predictive variable» Is the data biased? Censored? 25
Model Development Segmentation.080 Segment 1 Total Applicant Population Segment 2 Lowest Claim Frequency.068.066.064.062.060.058.056.054.052.050 Scorecard 1 Segment 1.1 Scorecard 2 Segment 1.2 Scorecard 3 Segment 2.1 Scorecard 4 Segment 2.2 0 1 2 3 4 5 6 # Scorecards... how many scorecards to deploy: Which segments matter most? Where the ROI is justified? How to consistently scale across all cards? 26
Model Development Score Engineering» Imposing constraints on the score formula to achieve various objectives» Objectives» Palatability» Legal requirements» Robustness over changing times» Adjustments for known sample biases» Engineering techniques» Choose the right objective function possibly more than one» Individual weight restrictions» Characteristic restrictions» Scaling across multiple models 27 1996 Fair, Isaac and Co, Inc.
Forecast model performance in production 28
Implementation Can I make it work in practice? Modeling Process Variables Scoring X Y Z Data Preparation Stats Modeling Data Stage Compare Modeling Repository Stats x y z Data Mapping Variables Scoring Deploy Runtime Package (Java) Production Data IT Model Deployment Process 29
Agenda Next Steps 30 2010 Fair Isaac Corporation. 2010 Fair Isaac Corporation. Confidential. Confidential.
Predictive Analytics: Achieving Greater Decision Accuracy» Capitalizing on predictive analytics» Consider and establish value of predictive analytics to achieving strategic goals» Develop focused strategy toward specifically targeted areas and needs» Identify partners and/or tools necessary to achieve strategic goals 31
Q&A Lamont D. Boyd, CPCU, AIM Insurance Market Director FICO Scoring Solutions LamontBoyd@FICO.com 602-485-9858 Date, 2010 32 This presentation is provided for the recipient only and cannot be reproduced or shared without Fair Isaac Corporation's express consent.