Detecting Fraudulent Claims for a Large Auto Service Provider
THE BACKGROUND
The largest auto service provider in the U.S. receives millions of invoices and claims per year for glass repair, replacement, and calibration services performed on customer vehicles. These invoices and claims are submitted by client-affiliated shops as well as 3rd party shops. They have detected potential fraudulent claims during manual review and observed that most fraud is committed by third-party shops by charging for services or products that are not required, pricing more than what is appropriate, or by submitting bogus claims when a vehicle has not been serviced. Reviewing each claim manually to check for fraudulent activity is not feasible. The client had approximately 3.2 million records per year of claim level data which captures granular level information of each claim such as shop and its information, exact job done, parts used, vehicle details, and customer information in a well-defined and structured format. The client wanted to build an A.I.-based solution to identify shops with unusual behavior that make these shops mindful of carrying out fraudulent activities.