Predictive Analytics

Our Expertise

There is a growing recognition within the insurance industry that traditional actuarial techniques, generally fall short of more sophisticated methods that involve statistics and machine learning. By bringing these advanced techniques to bear on actuarial problems, we are able to provide more accurate solutions and better insight into the cost drivers behind loss ratios.

Our actuaries are Certified Specialists in Predictive Analytics (CSPAs), and recognized experts in the industry. We have developed and reviewed hundreds of models, from standard Generalized Linear Models (GLMs) to Artificial Neural Networks (ANNs) and everything in between.

With our expertise, we have solutions to many problems. For example, it is common that companies using GLMs for rate models generate large swings in premium for individual insureds. However, manually dampening the proposed rate changes can bias results and lead to poor performance. We propose solutions to automatically identify the most important changes to make to rating factors. By doing so, we maintain a profitable and competitive book of business without clumsy manual adjustments that can lead to inaccurate rates and dire financial consequences.

Predictive Analytics Services

The following is a list of some of the many predictive analytics services we offer, with a brief description of each:

  • Statistical reserving models: More accurate reserves with better business intelligence about which segments are generating more claims.
  • Individual claim reserving: Help to focus a claims adjuster’s attention to the most important areas by determining reserves at the individual claim level,
  • Statistical reinsurance evaluation: Evaluate reinsurance offers to enhance negotiations and determine the optimal retention for the price.
  • Main effects rate models: Use GLMs to develop a rating model to improve underwriting performance.
  • Update existing rate models: Use advanced statistical techniques to reap the benefits of a next generation rating model without causing excessive policyholder dislocation.
  • Scoring models: Develop proprietary scoring models to enhance rating procedures.
  • Territorial models (clustering, geospatial, geodemographic): Analyze territories using a variety of advanced techniques.
  • Telematics: Enhance automobile models with telematics to optimally predict the riskiest driving behaviors.
  • Data mining models: We can assist in developing models to mine for valuable data that could improve company processes. This might include text mining or even image recognition algorithms.
  • Underwriting models: Model underwriting results to figure out which segments predict the best underwriting results.
  • Model evaluation / assistance with filing issues: We are experts in regulatory review of statistical models. We can assist with obtaining regulatory approval of rating models by evaluating the model, identifying major issues, recommending easy solutions, and developing materials to clearly communicate model assumptions to ensure quick approval.
  • Catastrophe modeling: We can assist in the development of stochastic catastrophe models for some catastrophes.
  • Other advanced modeling techniques for any of the above purposes (e.g., clustering, dimensionality reduction, splines, regularization, tree-based models, neural networks, etc.)

Find out what we can do for you