Big Data: Start small for big possibilities September 15, 2014
Big Data Agenda Intro to Big Data & definitions Example of traditional analytics Examples of Big Data analytics Big Data for insurance Call to action 2
Big Data Introduction Data drives decisions. And often dollars. Traditionally, data is structured in spreadsheets and databases and analyzed at specified time intervals. Traditional data management and analysis techniques are straining to adapt to less structured data, arriving in massive volumes, and needing real-time analysis. Welcome to the world of Big Data 3
Big Data In the insurance industry Insurance has always been one of the most data intensive industries Huge databases for: Marketing Policies and customers Historical Actuarial analysis Claims files but Big Data is different 4
Big Data Definition The 3 Vs that are characteristic of Big Data Data captured 2013 = 4.4 trillion GB Internet = 25 million GB Volume Velocity Streaming Data Real-time sensors Variety numbers, text, images, video, sounds, flat files, relational databases Big Data 5
Big Data The new technology ecosystem Hadoop and Map / Reduce NoSQL databases In Memory Analytics The Cloud Data Visualization Tools 6
Traditional analytics Geographic mapping of values, premium, etc. 7
Traditional analytics Geographic mapping of rates per $100 of value 8
Traditional analytics Compare modeled losses to premiums 9
Traditional analytics with more data Premium (Non-cat loss + Expenses + Net Cat Loss + Reinsurance) = PROFIT 10
Traditional analytics with more data Compare profitability with market share and agencies 11
Big Data Hail claims analysis View portfolio as of Aug 20, 2013 with hail storm 12
Big Data Hail claims analysis Compare claims with hail footprint Significant number of claims in localized area outside of hail footprint. Potential fraud? Some policies in the path of large hail but no claims filed. 13
Big Data Telematics Telematics the tracking of vehicle generated metrics in near real-time Data packets of vehicle metrics are typically sent to a central repository once every 15 to 60 seconds while a vehicle is operating Data can include items such as location, speed, acceleration (in all axes), engine performance, etc. Huge data management task 14
Understanding Driving Behavior Risk Driving Accident Fatalities 2012, NJ Heat Map 589 NJ Driving Fatalities (2012) 33,561 U.S.A. Driving Fatalities (2012) Source: Fatality Analysis Reporting System (FARS), Verisk Geospatial Concentration Analysis 15
Understanding Driving Behavior Risk Traffic Density Heat Map 117 Billion KM Travelled (2011) 63 Thousand KM of Road (2012) Sources: Federal Highway Administration, NJ Dept. of Transportation, TomTom, Verisk Analysis 16
Understanding Driving Behavior Risk ISO Risk Analyzer Model Loss Costs Heat Map $71 Min. Base Loss Cost $640 Max. Base Loss Cost Source: ISO Risk Analyzer Environmental Module, Personal Auto Bodily Injury Liability 17
GeoMetric Trip Analysis Source: Actual Driving Data Collected from ISO fleet 18
Driving Behavior Data 3.27 Gal / fuel 256.6 F 4200 RPM 72,852 Miles Dr. Seatbelt: Y 101 F 25 Mil Vis Wind: 2mph NW Sunny 2013/08/18 22:47:53.07 UTC 34 59 20-106 36 52-9.8 m /s 2 Interstate 40 (Freeway) Speed Limit 65 MPH Albuquerque, New Mexico 19
UBI Shows Unique Lift Beyond Traditional Proxies 3.500 Separation of Risks by Decile Using Safety Scoring Decile Loss Ratio Indexed to Average Loss Ratio (Actuals) 3.000 2.500 2.000 1.500 1.000 0.500 Actual loss ratio is 20% of the average These risks are overpriced smoothed curve Actual loss ratio is 225% of the average These risks are underpriced 0.000 1 2 3 4 5 6 7 8 9 10 Score Decile Range Safety Scoring is able to separate risks even after traditional rating variables have been applied Source: Verisk data and analysis (n = 3,024) ; Fifth decile has been replaced with average of fourth and sixth to adjust for anomalous data point 20
How Is Big Data Being Used in Insurance? Broker/Agent management Automated underwriting Retention assessment and management Early fraud detection Product Development Client Acquisition Underwriting Policy Admin Account Servicing Flexible and rapid market entry/exit Response uptake predictions Business quoting filters Risk prediction and rate setting Enhanced claims process management Subrogation/ recovery potential 21
Big Data In the insurance industry Implementation lags behind interest Even among larger companies, only 17% have incorporated Big Data into their modeling efforts. Big Data is not on the radar for most smaller companies 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Use of Big Data by company size 62% 8% 23% 7% <$1B 37% 12% 34% 17% >$1B Doing nothing with Big Data Would like to take advantage of Big Data, but it is cost prohibitive Evaluating/ implementing the use of Big Data Use Big Data and have incorporated it in our models Source: Earnix / ISO 2013 Predictive Modeling Survey
Big Data What does it mean for insurance? New Technologies to Understand New Skills to Learn New Business Processes to Design New Management Procedures to Implement New Opportunities to Seize But, ultimately, new Big Data technologies are just another tool in the toolbox for solving business challenges 23
Big Data Next steps Start saving data streams now Start with a small project that solves a specific business issue Continue to develop capabilities Build on successes Firms that adopt Big Data early will have a competitive advantage! 24
Thank You Christopher Nicolai Senior Vice President, Reinsurance Willis Re Phil Hatfield Modeling Data Services Executive ISO 25