A widely held notion in the industry today stipulates the use of advanced analytics allows for better claims decision-making and any results need to be automated if bottom-line improvements are to be achieved. As insurance companies start to buy and build different business analytical models for fraud detection, adjuster assignment, material damage appraisal, etc., these models must be incorporated smoothly into the claims operation. However, there is a danger of building multiple systems for managing and delivering model results.

For example, auto insurers may have a need to model the likelihood a material damage estimate is overwritten and get that information to a reinspector so they can select the higher-probability estimates for review. Simultaneously, they may have a need to detect both internal and external fraud on claims payments and transfer that information to a special investigative unit examiner. And finally, they want to be able to determine which customers are more likely to be dissatisfied with service and have a manager call those customers to smooth over any problems.

Clearly, all of these scenarios require different data sets and different models. The common mistake would be to develop or buy different systems to handle each of these challenges. There is, however, a technological solution–a model processing engine (MPE)–which should be considered by all insurers that embark on building, using, and automating multiple models for multiple uses. Simply put, an MPE is the quarterback for the input of necessary data for modeling and the output of the results the model produces to the appropriate automated workflow.