What should the claims management sector consider before jumping on the predictive modeling band wagon?

Predictive modeling has been used by the insurance industry for almost a decade. It has seen success primarily in pricing, underwriting, and in the development of new business through the micro-segmentation of existing risk pools.

Claims management appears to be the next segment of the insurance industry about to make the leap into the predictive modeling pool. Author Jim Kinzie noted in a recent article in Claims Magazine:

After more than a decade of observing underwriters using predictive models to improve results, many claim organizations have now embraced the notion that data can tell a very insightful story. By combining statistical analysis with the deep knowledge of claim experts, leading-edge claim operations have made data analytics the engine that powers their claim segmentation strategies.”  (Kinzie, 2010)

Before making the predictive model leap, the claims management authority within any company needs to consider both the advantages and disadvantages inherent in making the switch. The advantages are obvious – predictive modeling offers the potential to enhance the insurer’s bottom line through a number of mechanisms. Better detection of fraudulent claims, more efficient allocation of resources to claims with the highest potential payouts, flexibility in monitoring the claims process at all points in that process, and increased customer satisfaction because of streamlined claims handling procedures.

The disadvantages of switching to predictive modeling in the claims management organization are less apparent at first glance, but are none the less real. In a white paper on predictive analytics,  Senior Director of Knowledge Research for the AICPCU/Insurance Institute of America, Charles Nyce, noted some of the disadvantages and problems associated with predictive modeling:

• Inherent inaccuracy of the predictive model
• Cost of implementing predictive analytic techniques
• Resistance to change within the organization
• Need for clean, accurate data  

Predictive model accuracy is the first element in Dr. Nyce’s list (Nyce, 2007). He is not alone in voicing concern about the accuracy of predictive models. For example, in an article in Information Management Online, Senior Director of Knowledge Resources and ethics counsel at the AICPCU/Insurance Institute of America, Donna Popov, says definitively – “A model’s potential for inaccuracy is an important consideration for insurers that rely on predictive modeling. Just like a credit score, a predictive model indicates what is likely to occur, not what is certain to occur.” (Popov, 2009)

In the section on Data Availability in the same white paper, Dr. Nyce keys in on the second element in the list above – that of cost. He notes that while data availability is usually not a problem for insurers, the data may be stored on legacy systems which are incompatible with systems used for predictive modeling. “Converting data on these legacy systems to a usable format can be time consuming and costly,” notes the author. (Nyce, 2007)

While Dr. Nyce, in his third point, focused on “Resistance to change” from the employee base in the organization, other experts have noted that one of the pitfalls of moving to predictive modeling is more management than employee focused (Nyce, 2007). In a presentation before the Casualty Actuarial Society, senior consultant for EMB America Jeff Kucera, emphasized that senior management, “May not understand cultural changes to be made, e.g. some risks previously considered uninsurable may now be insurable,” and that management may, “Still have favorite risks that they want to heavily discount.” (Kucera, 2005)

The need for “clean, accurate data” is a pre-requisite for a successful transition to a predictive modeling based claim management system. In his list, Dr. Nyce noted that that requirement is often problematic (Nyce, 2007). That observation is echoed by insurance blogger and author, Joe McKendrick. In a recent article in Insurance Networking News, McKendrick observes that, “…for many insurers, an ability to get to the right data – or poor data quality – is hampering their ability to achieve such capabilities.”  (McKendrick, 2010)

There can be little doubt that predictive modeling is making steady inroads into the claims management operations of many leading insurers. However for any given insurer considering the move to predictive modeling, there remain considerations that are specific to the circumstances of that insurer. These individual circumstances will dictate whether, or if, the time is right to make the move.

References:

Kinzie, Jim. (6/2/2010). Driving Bodily Injury Claims. Claims Magazine
Retrieved from http://www.claimsmag.com/Issues/2010/JUNE-2010/Pages/Divining-Bodily-Injury-Claims.aspx?k=Divining+Bodily+Injury+Claims

Kucera, Jeff. (2005). Predictive Models, Pitfalls and Potentials. Casualty Actuarial Society
Retrieved from http://www.casact.org/education/annual/2005/handouts/kucera.pdf

McKendrick, Joe. (7/1/2010). The Path to Analytics is Paved With Good Data. Insurance Networking News. Retrieved from http://www.insurancenetworking.com/issues/13_8/insurance_technology_
data_management_analytics_warehousing_MDM-25150-1.html?pg=1

Nyce, Charles. (2007). Predictive Analytics Whitepaper. AICPU Insurance Institute of America
Retrieved from http://www.aicpcu.org/doc/PredictiveModelingWhitePaper.pdf

Popow, Donna. (5/15/2009). Models Are Not Reality. Information Management Online
Retrieved from http://www.information-management.com/news/predictive_modeling_claim_
insurance-10015433-1.html

(CC) August 2010 Vatti-Manhattan Group
This article may be copied and distributed freely provided that you keep this copyright notice intact and is provided for free and without charge.  We have to the best of our knowledge abided by all copyrights, trademarks and quoted material.  Any comments, omissions, misuse concerns etc. regarding the use of trademarks and copyrights should be directed to
info@vatti-manhattangroup.com.