More Insurers Adopting Predictive Analytics to Combat Fraud

Declining cost barriers, the availability of cloud storage and telematics will likely increase predictive analytic adoption rates, according to Dan Donovan, assistant vice president of Claims Solutions at Verisk Insurance Solutions.

Though predictive modeling has been used for some time to thwart fraud, he said, it was mainly used by larger carriers.

“A lot of that had to do with the cost of predictive analytics, the lack of data science teams that existed within the companies themselves…the IT requirements and cost to support analytics in large data environments,” Donovan explained.

In those instances, he said, custom models were built specifically for a carrier’s dataset. At the time, the models needed to be hosted on the carrier’s premises.

“Carriers were very reluctant to let any data outside of their own firewall that might contain personally identifiable information or personal health information,” said Donovan.

The other issue, he said, was that initially insurers didn’t understand predictive analytics and how to operationalize the output.

Over time, as cost barriers declined and insurers hired data scientists and began using the cloud, insurers could see the benefit of utilizing predictive analytics affordably.

The types of fraud that predictive modeling can identify and predict is limitless, he said, noting that it’s anything that data can be applied to, starting from the from point of sale.

“You can start attempting to identify fraud at the point of sale, when that application is first submitted to an agent through the underwriting process. We’ve even had discussions with carriers about looking at agency books of business,” Donovan said. “Really, it can start there and then move onto when a claim occurs and then after claims and at the point of renewal for policies, you could again evaluate.”

In addition, any book of business can benefit from predictive analytics and fraud scoring, he said.

“From a claim standpoint, it’s limitless to the lines of business and the coverage types that you can apply analytics to,” Donovan said.

Predictive analytics have been applied to some lines of insurance more easily than others.

For example, auto physical damage and bodily injury has seen the greatest uptake in predictive analytics, Donovan said.

“There’s been an increase in interest in homeowners’ property claims for fraud, but that’s been very slow on the uptake,” he said.
The use of the Internet of Things (IoT) and telematics is a game changer to combat fraud, said Donovan.

“IoT’s extremely interesting to the industry as a whole, not just in the fraud space…we really see this as a game changer and believe it represents the next generation of fraud detection capabilities,” Donovan said. “When you start to be able to bring in telematics and IoT data from smart homes and smart auto into the claims process, into the underwriting process, it’s really going to change the game.”

Both offer data that will likely deter soft fraud.

“These claims where you’ve got exaggerated injuries or inflated medical treatment, where the accident may have legitimately happened, but the damage was enhanced or intentional,” Donovan said. “It’s very difficult to prove fraud in these types of losses. Our belief is that for carriers who can start to gather and incorporate IoT and telematics data into their claims and underwriting process, this is really going to allow them to flip that over and be more objective in their evaluations of losses.”

Wearables may impact fraud even more, he said, describing a case where a heart monitor’s data was used to prove an individual may have been involved in an arson. According to an Associated Press report, police say data recorded by a man’s cardiac pacemaker helped lead to his indictment on charges of aggravated arson and insurance fraud in a fire at his Ohio home. Police say the man gave statements inconsistent with evidence obtained during the investigation. They say he told them that he packed some belongings when he saw the fire, threw them out of a window and then carried them to his car. Court records show a cardiologist reviewing the man’s pacemaker data said his medical condition made it “highly improbable” to have taken all of the actions he described.

“The deterrent effect may ultimately end up being the biggest thing that IoT and telematics does from a fraud standpoint, when people start to realize that all of their moves are being tracked, that you can pull information from their car or from their home that will either support their version of events or completely turn it upside down,” said Donovan. “That may get people thinking twice about filing fraudulent claims.”

The Associated Press contributed to this article.