Interesting news on discussion of predictive analytics at the Casualty Actuarial Society (CAS) - PREDICTIVE MODELING RAISES OPPORTUNITIES AND ISSUES FOR ACTUARIES AND INSURERS, CAS ANNUAL MEETING IS TOLD. Typical quote:
"The use of predictive modeling techniques is becoming increasingly popular among a wide variety of insurance companies because of its potential to give insurers – especially personal lines insurers – an edge in competitive markets."
A number of key issues were identified at this meeting:
- Don't blindly implement - make sure you understand the impact
- Explicability - make sure you can explain how the models work to a regulator
- Right tools - use appropriate tools
An Enterprise Decision Management approach works perfectly for this. Business rules and scorecards make it easy to explain decisions. Unlike other kinds of predictive models, Fair Isaac scorecards are specifically designed to show how and why a particular score resulted. Business rules are also easy to understand and make it easy to track which rules fired for a given decision. Taken together they can make automated decision, enhanced by predictive analytics, easy to explain and justify.
The value of these analytics is in the automation of decisions. Thus the right tool not only has to help build the models, it must make it easy to implement them in an information system reliably. This is what Fair Isaac's Model Builder technology has been designed to do. Working with a company, like Fair Isaac, that has experience building insurance models is also a good way to go.
Dislocation of staff was also noted as a risk although our experience that using predictive analytics, at least in underwriting, results in the ability to focus staff on higher value activities, such as agency management and "book of business" analysis, instead of the mechanics of underwriting an individual policy.