Client
· A Pharmacy Benefits Manager
· There are about 4 billion prescriptions written is US alone every year.
· FBI estimates, potentially, up to about 10% of the transactions can be classified as FWA
Challenge:
· FWA detection
· Current process is manual, can only find known fraud signatures
· Lots of false positives and undetected fraud (fraudsters are always one step ahead)
Action:
· They invested in our Jumpstart Program.
· We got several years of transaction data in our lab. Our data scientists went to work
· We developed serval supervised and unsupervised learning algorithms to identify FWA
· Only 16 weeks, from start to finish
Outcome:
· Potential for elimination of manual work with full automation
· 80% reduction in false positives, ability to find previously undetected fraud
