Advancing Business Value Through Predictive Analytics and Data Science
2 likes284 views
The document discusses how The Hanover Insurance Group is advancing business value through predictive analytics and data science. It outlines top analytics trends like product innovation, improving the customer claims experience through behavioral analytics, and improving underwriting profitability with advanced analytics. The presentation also provides examples of how Hanover is using claims analytics for fraud prevention, claims triaging, and leveraging analytics to power faster responses to catastrophic events.
Advancing Business Value Through Predictive Analytics and Data Science
- 1. 1 Srinivasan Sankar The Hanover Insurance Group Advancing Business Value Through Predictive Analytics and Data Science Boston CDO Executive Summit - June 26, 2018 - Hyatt Regency Boston
- 2. 2 Business Overview • National P&C Insurance company founded in 1852 and headquartered in Worcester, Massachusetts • Traded on NYSE under ticker symbol THG • One of the largest insurance companies in the U.S. • Approximately 5,000 employees across the U.S. & U.K. • Named by Forbes magazine as one of “America’s 50 Most Trustworthy Financial Companies” • Partners with a select group of approximately 2,200 independent agencies nationwide for product distribution • Underwrites a wide range of personal and commercial lines for businesses, individuals, and families • Integrated several strategic U.S. and international acquisitions in recent years that dramatically expanded the company’s product portfolio and geographic footprint • $5B+ in revenues • Excellent rating from the major ratings agencies
- 3. 3 Disclaimer This presentation is for general informational and discussion purposes only. The opinions expressed in this presentation are those of the presenter and not necessarily those of The Hanover Insurance Group (“Hanover”). Neither Hanover nor the presenter guarantees the accuracy or reliability of the information provided herein. No representation or warranty, express or implied, is provided by the presenter or Hanover in relation to the fairness, accuracy, correctness, completeness or reliability of the information, opinions or conclusions expressed herein. The information contained in this presentation is subject to change without notice. This presentation does not, and is not designed to, provide legal advice and meeting participants should consult with an attorney concerning the use of data to ensure all legal and regulatory requirements are satisfied.
- 4. 4 Agenda • Top Five Analytics Trends • Shift from traditional analytics to modern analytics • Old World vs. New World • Claims Analytics – Fraud prevention • Claims Triaging, Claims Mitigation • Analytics powering CAT response • Questions and Comments
- 5. 5 Innovation in P&C Insurance Top Five Analytics Trends 1. Product Innovation on the Rise and Product Analytics Are Required – Launching new products based upon emerging needs of digital and millennial customers 2. Adopting New Types of Analytics to Improve the Customer Claims Experience 3. Building Behavioral Analytics Using IoT Data and Applying It Throughout The Enterprise – Location, Time of Day, Safety Indicators, Driving Patterns 4. Improving Underwriting Profitability Through New Data and Advanced Analytics – Embedded Analytics Using Big Data and Artificial Intelligence – Automated Underwriting 5. Fraud Analytics
- 6. 6© Gartner Research, Inc. By 2021, 75% of prebuilt reports will be replaced with or augmented by automated insights delivered on a "most needed" basis. Smart Data Discovery Visual Exploration Analytic Dashboards Storytelling/Narrative Traditional analytics platforms are designed to support modular development of IT- produced analytic content, specialized tools and skills, and significant upfront data modeling, coupled with a predefined metadata layer, is required to access their analytic capabilities. A modern analytics platform supports IT-enabled analytic content development. It is defined by a self-contained architecture that enables nontechnical users to autonomously execute full- spectrum analytic workflows from data access, ingestion and preparation to interactive analysis, and the collaborative sharing of insights. Traditional Analytics to Modern Analytics - A Strategic Outlook
- 7. 7 How the shift happens – some examples Old World New World • Product recommendation from advisor experience • Risk group classified using static data, e.g. age and medical history • Labor-intensive claim processes take months to finish • Customer churn analysis after the fact • Many loopholes in insurance fraud detection • Product recommendation from deep analytics of customer data • Risk group classified using broad spectrum of data, e.g. social media, click streams and web analytics • Claims processed faster using automated image classification • Customer churns predicted and intercepted in real time • Insurance frauds more accurately detected from customer 360º analysis • Increased revenue
- 8. 8 Claims Analytics – Fraud Prevention Loss costs are the key components of insurer results. Fraud prevention is $36B in the bottom-line Loss $6k Expense $3k Profit $1k Premium $10k Identifying Fraud here Will increase the bottom-line here Claims Fraud Model Segment and Prioritize Claims Identify Fraud Potential Rare event modeling Algorithms SMOTE* (to handle skewed Data) Cost Savings • 10% of Total Loss goes into Fraud • $600 B P&C industry • Loss Ratio @ 60% avg. • Total loss = $360B • 10% = $36B *Synthetic Minority Over-sampling Technique Fraud at a glance
- 9. 9 • Long tail claims (bodily injury claim severity escalation) • Predictions based on historical data • Surgery predictions • Attorney involvement prediction • Input to Claim mitigation in the business process Claims details EDW Unstructured claim notes Claims System Machine Learning Deep Learning 15+ Algorithms for Fraud Detection Modeling and Analytics Real-time Scores Green = Procced further in the Claim Process Yellow = Proceed with Caution Red = Needs further investigation Fraud prevention, Real-time Scoring and Predictive Analytics Claims Triaging, Claims Mitigation
- 10. 10 Analytics powering faster CAT response • Storms tracking - keeping watch on where and when landfall expected • Potential coverage impacts (Flood, Building, Marine, Auto, etc.) • Help Claims department in determining how to act for each state so they could mobilize and deploy staff