Insurance & Technology is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Claims

05:21 PM
Connect Directly
Facebook
Google+
Twitter
RSS
E-Mail
50%
50%

Belief Network Model Helps AIGM Reduce Leakage

The insurance industry has made a substantial investment in fraud detection, but simple overpayment of claims still results in substantial leakage suffered by insurance companies. Predictive modeling can help stem that leakage by identifying claims with a high probability of overpayment.

The insurance industry has made a substantial investment in fraud detection, but simple overpayment of claims still results in substantial leakage suffered by insurance companies. Predictive modeling can help stem that leakage by identifying claims with a high probability of overpayment.

In addition to the many sources of claims adjuster error, about a third of insurance customers are willing to accept an overstated claim, according to Frank Cacchione, head of New York-based PA Consulting's insurance practice. "The more the industry invests in overcoming errors in claims management, the more it is benefiting not only its own bottom line, but also its customers who are not predisposed to lie, cheat, steal or allow mistakes to be made in their favor," he comments.

PA Consulting helped AIG (New York, more than $600 billion in assets) subsidiary and personal lines auto and homeowners insurance marketer AIG Marketing, Inc. (AIGM), to invest in a claims predictive modeling solution (CPMS) in Fall 2002, built on Bayesian belief network models, which analyze likely outcomes of events.

The first step of the solution was to analyze the claims adjustment process to identify areas that allowed for leakage, according to Cacchione. Then relevant data was gathered from disparate sources into a single, purpose-built database. Third, Bayesian belief network models were constructed to analyze the relationships between the different data points of every claim. "Finally, we brought all of this together into a system that would allow the material damage organization to re-inspect appraisals that carry a high probability of being overwritten," Cacchione explains.

Field estimates from AIGM adjusters or preferred shops are uploaded to the company on a same-day basis. As the information arrives, it is passed through the model in real time. "The model takes the claim and evaluates all kinds of information, such as policyholder information, bodily injury data, etc., and comes out with a probability level of whether the settlement was too high," remarks Cacchione. If that probability is high, "Re-inspectors go out, generally the very next day, look at those vehicles, and override the original claim, if appropriate."

The Bayesian methodology works in self-learning fashion, so the CPMS -- currently in the process of being rolled out countrywide -- gets cumulatively better. But the system also helps AIGM's human partners to learn. "It helps them to focus on where errors are coming from and turn that into corrective models for their staff and preferred shops," Cacchione says. "And since independent appraisers and repair shops know that AIGM has an improved procedure for adjustment re-inspection, they are writing even tighter estimates."

Anthony O'Donnell has covered technology in the insurance industry since 2000, when he joined the editorial staff of Insurance & Technology. As an editor and reporter for I&T and the InformationWeek Financial Services of TechWeb he has written on all areas of information ... View Full Bio

Register for Insurance & Technology Newsletters
Slideshows
Video