1555 track 1 huang_using his mac
0 likes115 views
1) The document discusses challenges in predictive analytics projects including finding ideas, assessing impact, prioritizing projects, developing and deploying models, and post-implementation monitoring. 2) It describes moving from manual processes to automating workflows to using decision-making and machine learning to drive innovation. New data sources like prescriptions and driving records can help leverage new predictors for better customer experiences. 3) A framework is presented for evaluating analytics projects based on their value, data and process integration requirements, urgency, and ease of implementation to identify high priority projects that maximize return. The top priority projects identified were predictive underwriting, pricing models, and customer journey analytics.
1555 track 1 huang_using his mac
- 1. CREATING VALUE THROUGH ANALYTICS INNOVATION Wayne Huang, PhD Head of Predictive Analytics, Group Insurance Prudential Financial Inc. 1 - Predictive Analytics World, NYC, October 30, 2017
- 2. Analytics Team Challenges 1.⯠Find analytics ideas 2.⯠Assess solution impact 3.⯠Prioritize analytics projects 4.⯠Develop & deploy analytics models 5.⯠Post-implementation monitoring 2
- 3. From Automation To Innovation Manual Process Workflow Automation Decision & Learning Innovation Productivity Time 3
- 4. 1. Find The Diamonds 4 Underwriting Review High Risk? Order APS/ Med Exam Approved Issue Policy Denial Letter Insurance Application Underwriting Review Life Insurance Underwriting Process MIB Check YesNo Yes No
- 5. 2. Leverage New Data Old Paradigm §⯠Traditional data (Application, MIB, APS, Labs) §⯠Underwriter judgement §⯠Long decision time §⯠Higher cost §⯠Customer abandon application §⯠Painful customer experience New Paradigm §⯠New predictors (Rx, MVR, Medical, Credit,âŚ) §⯠Predictive model §⯠Decision automation §⯠Lower cost §⯠New revenue §⯠Superior customer experience 5
- 6. 3. Control Uncertainty Uncertainty is a result of having to deal with ambiguity and too many variables. Itâs a risk to business. 1.⯠Use analytics to reduce ambiguity §⯠Use campaign response predictive model to identify customers propensity §⯠Use cluster analysis for market segmentation, product positioning, and targeted campaign 2.⯠Use machine learning to identify and manage variables that have higher predictive power for better outcome §⯠Use pricing predictive model to assist insurance product pricing §⯠Use prospect scoring model to determine new customer handling priority 6
- 8. Operational Feasibility Assessment 1.⯠Understand the current process 2.⯠Identify tasks impacted by the solution 3.⯠Design a new process 4.⯠Assess the impact of solution implementation §âŻLabor â case handling volume change, new skill requirement §âŻInput â existing data vs. (real-time) new data §âŻTime â cycle time reduction §âŻTechnology â system change requirement §âŻLegal and compliance 8
- 9. Financial Viability Assessment 1.⯠What value can the analytics project bring? §âŻLower cost, shorter cycle time, more revenue, higher profit 2.⯠Whatâs the model development and implementation cost? 3.⯠How to quantify the value? §âŻCost Savings: unit labor cost x reduced volume by automation, plus any other input reduction x unit cost, minus new data cost §âŻNew Profit: new sales x profit margin, or new customer acquisition x life time value 4.⯠Calculate five year NPV of the investment Year 1 Year 2 Year 3 Year 4 Year 5 Total Cashflow -$1MM $500K $1MM $1MM $1MM NPV (10% ROE) -$1MM $413K $751K $683K $621K $1.5MM 9
- 10. Recognize Indirect Value Activities 1.⯠Activities that generate value to customers and firm §⯠Primary activities in value chain §⯠Predictive underwriting can reduce cost, increase revenue and improve customer experience 2.⯠Activities that preserve value §⯠If neglected, a firm loses the ability to generate economic value, e.g., advertising, customer service §⯠Targeted campaign can increase new customer acquisition and reduce cost §⯠Mining customer support data can identify product improvement opportunities and increase sales 3.⯠Activities that enable options (deferred value) §⯠Give a firm an advantage in dealing with uncertainty and change. They help exploit value-generating opportunities, e.g., R&D, employee training §⯠Predictive analytics can help speed up life science and material science discovery §⯠Predicting training effectiveness can better prepare employees for new tasks 10
- 12. Why Analytics Projects Failed 1.⯠Lack of alignment with strategic goals §⯠Inability to integrate analytics solutions into workflows §⯠Limited front-line adoption 2.⯠Lack of strategic alignment and direction §⯠Lack of stakeholder alignment or support §⯠Lack of clear roadmap 3.⯠Poor data quality §⯠Missing or incomplete data §⯠Data quality or accuracy issues §⯠Data fragmentation 4.⯠Other §⯠Missing team skills or capabilities §⯠Unclear use case scope §⯠Inability to articulate value McKinsey & Company, âRaising Returns on Analytics Investments In Insuranceâ, 2017 12
- 13. Project Evaluation Framework Project Value (5 Year NPV) Analytics Solutions Data Availability Process Integration Urgency Project Ease* 1. Pricing Predictive Model $3M Complex Partial Medium Difficulty Normal 4 2. Claims Analytics $2M Medium Complexity All Medium Difficulty Time Critical 3 3. Customer Journey Analytics $1M Simple All Easy Normal 1 4. Predictive Underwriting $3M Medium Complexity Partial Medium Difficulty Normal 3 5. Prospect Prioritization $1M Medium Complexity Partial Difficult Time Critical 4 6. Campaign Response Model $1M Simple All Easy Normal 2 13 * Project Ease ranges from 1 to 5, with 1 being the easiest and 5 being the most difficult.
- 14. Ă⯠The goal of the heuristic is to choose projects, which maximize the ratio of NPV to Project Ease Ă⯠Suppose the total budget is $2.5M Project Selection Heuristic Project Cost Value Ease Ratio Strategic* Rank 1. Pricing Predictive Model $500K $3M 4 .75 1.13* 2 2. Claims Analytics $500K $2M 3 .66 .66 4 3. Customer Journey Analytics $250K $1M 1 1 1 3 4. Predictive Underwriting $1M $3M 3 1 1.5* 1 5. Prospect Prioritization $500K $1M 4 .25 .38* 6 6. Campaign Response Model $500K $1M 2 .50 .50 5 14 Ă⯠Select Projects 4, 1, 3, 2 for a total cost of $2.25M * Strategic projects receive 50% additional weight