Articles

---

Predictive Modeling for the Insurance Industry

Life Insurance Actuary to Claim Scoring Geek

By Barry Senensky

Ten years ago I would never have imagined that I would ever write an article about disability claims. I had been a practicing actuary since 1979 with a career focus virtually exclusively on financial reporting and asset liability management and no real interest or desire in changing.

But during a "what-have-you-been-up-to-since-we-last-talked" business lunch with Jonathan Polon, a former actuarial colleague of mine turned predictive modeler, I listened fascinated as Jonathan relayed to me his exploits into the field of data mining and predictive modeling. I was soon convinced of the incredible power of these tools for decision making.

The career-changing thought then occurred to us that there was a great opportunity to bring this predictive modeling technology to the life insurance industry – and who better to do so than a couple of actuaries with life insurance experience? Before that lunch was over, Claim Analytics Inc. was born. We would combine our actuarial strengths with predictive modeling expertise to bring provide predictive modeling to the life insurance industry.

Soon after our lunch meeting, we zeroed in on the significant potential for predictive modeling in the disability insurance market. We began to develop a predictive model to score disability claimants, based on their likelihood to recover in a given timeframe. The model combined the latest in computing power, advanced pattern detection technology, and Jonathan's experience in predictive modeling.

The completed model gave each claimant a score between 1 to 10, based on their historical claims data. A score of 1 indicated a 0-10% probability of return to work within the given timeframe (generally 6 months or 24 months) a score of 2 a 10-20% probability of return to work within the given timeframe, a score of 3 a 20-30% probability of return to work within the given timeframe, and so on.

Our validation and testing processes on our model generated very good results. We now had a validated claim scoring tool which objectively and successfully assessed a claimant's likelihood to recover. What a boon to the claims management process! We naively thought we would have customers lining up at the door to buy our ground-breaking product.

It is now ten years later and Claim Analytics Inc. is well established in providing predictive modeling to the insurance industry, with several major North American insurers as clients. Our product offering includes not only our disability scoring product, but pricing, reserving and fraud detection tools as well. Yet while there are now many companies that use objective claim scoring as a management tool, we have come to learn that wide-spread acceptance of a new scientific technology happens quite a bit more slowly than what we originally anticipated.

However, instead of being discouraged by this, I am more confident than ever that automated claims scoring will become commonplace and an industry standard. Automated objective, consistent and accurate claim scoring just makes so much sense! Especially in an industry with a foundation of probability and statistics. An industry that is run based on objectivity, science and hard numbers cannot ignore the tremendous benefits that this type of capability offers.

Our customers have found that automated claims scoring delivers a number of critical benefits to disability claims management. Things that I have directly observed with our clients are that it:

• helps them quickly establish a starting point for a new claim
• helps balance case workload
• facilitates early intervention
• helps optimize resource allocation including the deployment and use of rehabilitation and external resources.

As an overall management tool, it allows case managers to be realistic about the use of their own time and energy. The scores quickly highlight cases that are most likely to benefit from their experience and efforts, and perhaps more importantly, to let go of cases where the return for their time and effort is likely to be low. All this results in improved financial performance.

While we haven't had customers "lined up at the door" as we so naively thought 10 years ago, I am extremely pleased with the progress Claim Analytics has made in what seems like a very short time. I have seen many examples of how helpful it is for our clients to be able to use their existing data to manage business performance going forward. I am even more pleased that my career took the turn that it did and I became involved with the development and evolution of such useful products …. predictive modeling products that I truly believe will become standard in the industry.