Willis Towers Watson to deliver real-time claims fraud detection using machine learning

Insurers are going through significant transformation driven by rising customer expectations and a new wave of digital-first competitors disrupting the market. Consumers now demand instant 24/7 services, combined with the ability to self-serve, benefit from fast-track claims processing, and receive unparalleled customer support.

“Radar Live works behind the scenes, enabling use of advanced analytics to assess structured and unstructured data to uncover hidden fraud patterns and complex relationships within social networks, giving us the capability to deliver scoring for fast review and accurate assessment” said Sascha Firle, Director Advanced Analytics at Trygg-Hansa. “Better claims fraud detection capabilities and faster response times help us to detect fraud early in the cycle and result in fewer false positives. This reduces pay-outs on fraudulent claims, leading to improved profitability and a better customer experience for genuine claimants.”

Based on initial tests run by Trygg-Hansa, Radar Live was able to increase the insurer’s fraud detection rate by more than 100 per cent. According to Willis Towers Watson, this significant performance lift will continue to improve as advanced analytics and machine learning models continue to learn the characteristics of fraudulent claims, are augmented with additional internal and external data, and adapt to identify new patterns.

"We are delighted to be working with Trygg-Hansa, recognised as one of the market’s most progressive, technology-driven insurers, as we help them respond to new threats and stay at the forefront of claims fraud detection,” said Tom Helm, Head of Claims Consulting at Willis Towers Watson. “Traditional methods are sub-optimal as they rely on slower overnight batch runs, fixed business rules, which only anticipate suspicious activity based on past fraud and may leave insurers dependent upon external providers to make any analytical improvements.

“To combat today’s complex and evolving schemes, Radar Live allows insurers to run multiple machine learning algorithms, assess a vast array of data points and produce a score on the likelihood of fraud all within less than a second. This enables decisions to be made, whilst the customer is on the phone or notifying their claim on-line. Insurers also remain in control of the analytics, enabling them to respond in an agile way and quickly deploy new fraud rules and models into production.”