Allstate and Nationwide among insurers using public data sources in fraud detection
The insurance industry is looking to digital technology to help lower claims costs in several ways, and that includes new approaches to fraud fighting.
Policyholder fraud continues to be a leading cause of incurred underwriting losses annually in the U.S. And carriers are now ramping up detection efforts to fight back. The Insurance Information Institute finds nearly 10% of annual property & casualty losses result from fraudulent claims data. A 2017 Verisk Analytics study also found that premium leakage—defined as missing or erroneous underwriting information—amounts to nearly $29 billion a year in losses for personal auto insurers.
Allstate is now a few months into its announced fraud detection partnership with insurtech Carpe Data in November. The deal gives the carrier’s claims adjusters license to request on-demand background checks on policyholders using publicly available information. Carpe Data then utilizes AI algorithms to scour social media sites, public databases and news media outlets for proof of limited physical prowess or recent criminal activity.
“It’s an additional investigative tool that helps us confirm information we’ve been told, but have not been able to validate through traditional means [taking statements, making phone calls or hiring an investigator,]” said George Naftzinger, director of third party claims at Allstate. “The alternative is to have an individual adjuster sifting through everything on Google themselves, which takes time.”
The Carpe Data partnership represents Allstate’s quickest approach to detecting fraud in less than 24 hours, Naftzinger says. The insurer is now weighing other potential fraud detection options, which it chose not to disclose.
Nationwide also partnered with Verisk Analytics to thwart policyholder fraud last month. The agreement enables the insurer to leverage Verisk’s ClaimsDirector software to inspect first notice of loss information as it comes in. Versik’s technology suite gives clients access to a business intelligence dashboard, an interactive prior loss timeline and triage page. Nationwide declined to comment for this story.
Policyholder privacy at risk
Privacy is the biggest concern resulting from carriers’ efforts to use public data to identify fraudulent claims, according to Jeff Goldberg, SVP of research & consulting at Novarica.
“It can be potentially problematic behavior for an insurer to review social media for deep analysis on every claim,” he said. “Even if you can, that will be frowned upon by regulators, social media companies and policyholders.”
Insurance companies have previously used variables such as credit scores as stand-ins for better, more useful data to determine risk. But with the amounts of available data growing exponentially in recent years, carriers now have direct access to proof of falsified claims. The problem is the data is less structured, and much harder to analyze because it’s not refined with a nice API, Goldberg says.
Using AI to fight fraud may also automate bias. AI algorithms generally make future predictions based on past analyses done by humans. The belief is AI will have such a large set of data to pull from that it will fix that. But bias can impact future decision making, especially in the beginning as the AI is learning. Another recurring problem with AI is that it’s harder to understand the decision it tells users to make.
“That black box can lead to the insurers trusting the output without digging into it,” said Goldberg, adding that telematics is the safest way to fight insurance fraud to date. “If two cars are involved in a fender bender and one person claims the other reversed into him, you [the carrier] would easily be able to tell.”
In the future, insurers can also leverage usage-based insurance programs as fraud detection tools, Goldberg says. However, UBI today is mainly utilized as a means for behavioral adjustment. Many carriers would also prefer not to accuse policyholders of fraud using information they opted to give in exchange premium discounts.
“The people lying to you aren’t the problem,” Goldberg concludes. “It’s just a smaller scale problem of the social media issue where you look at my data record specifically just to second guess me.”