About
Subscribe

Advanced analytics combats insurance fraud

By Cathleen O'Grady
Johannesburg, 06 Aug 2013

High-performance is an essential tool in detecting fraudulent insurance claims, says Tracy Dunbar, analytics director at BITanium.

An estimated 10% of insurance claims are fraudulent, says Dunbar, but more than 80% of these fraudulent claims are not detected as such, which amounts to millions lost annually. According to Dunbar, traditional techniques to detect fraud rely on a combination of rules hard-coded into the IT system and on the gut feel of the claims assessors, adding that these systems are insufficient to protect insurance providers from fraudulent claims.

"The rules are seldom reviewed and updated, and do not target new and/or changing patterns of fraud," she explains. "Fraudsters learn the rules and find loopholes. There is no business agility, as any modifications to the rules require substantial change requests and time delays to implement."

An additional problem with this approach is the need for qualified and experienced assessors, who are in short supply, she adds, with systems struggling as the amount of clients and volume of to be processed increases. She proposes high-performance analytics as a solution.

"Advanced analytics uses a combination of statistical and mathematical techniques and computing power to find patterns in data that are not evident using more traditional methods," Dunbar explains. "Advanced analytics empowers insurers to develop models to target changing profiles of fraud and to implement different models for each line of business."

High-performance analytics uses data from insurance providers' data warehouses, as well as external data, such as public records, to better identify fraudulent claims and predict potential risks. "Third-party data that can be valuable in contributing to fraud identification might include data on bankruptcies, litigation, address change velocity or even medical bill review data. A wealth of information also resides in unstructured sources, including social networks."

Dunbar adds that these models can be incorporated into a real-time solution, with rules to identify suspicious behaviour and how to investigate the best course of action for each claim.

With the right infrastructure and technology, evidence-based claims decisions can be and made in real time, saving time and improving detection, she says. "Potentially suspicious claims can be automatically directed to the forensics department at claim registration. In the near future, I suspect insurers will offer their clients a way to register their claims online and monitor the progress of their claims."

Advanced predictive analytics allows insurers to organise and analyse both structured and unstructured data; to deliver more accurate assessments of fraud risk; and even predict future fraud attempts, Dunbar concludes. "By streamlining fraud assessment, predictive analytics reduces fraud losses and also speeds up claims processing, which results in a better customer experience."

Share