Tide APP Fraud Model: Building Risk Controls at Onboarding

Saurav took on the work as UK APP fraud rules created a new liability exposure for payment providers. Tide needed a way to identify likely perpetrators at onboarding while keeping legitimate members moving through the product.

He defined a new-to-Tide risk product that scored each applicant's probability of committing APP fraud in the first six months. He worked with data science on the model proposition, including low, medium, and high risk bands built from roughly 45 to 50 behavioral, identity, device, and third-party signals.

Saurav owned how the model would operate in the product. High-risk members moved into enhanced due diligence, while medium-risk members received progressive access and could unlock higher limits by providing further documentation.

He drove alignment across engineering, operations, financial crime, quality assurance, threat modeling, and executive leadership. When a cloud migration threatened the launch schedule, he secured an exception so the work could continue before the regulatory deadline.

The team back-tested the model before launch and found it could identify up to 35% of fraud volume at onboarding. Over the following year, APP fraud volume fell from about £15 million to about £5.5 million.