Skip to main content
Solving the Retail Banking Fraud Challenge

While I frequently blog about analytic innovations in card fraud detection, protecting retail banking accounts is equally important—and equally challenging. Banks worldwide are recognizing the need to apply fraud analytics to demand deposit accounts (DDA), including internet banking, mobile banking and other non-card related electronic payment channels. In many regions, the amount of funds transferred through these channels far exceeds those typically seen on payment cards, yet fraud protections are often not as mature.

Protecting retail banking accounts, however, presents an analytics challenge. Given the newness of these electronic payment channels for many banks, there is often a lack of historical and fraud data on which to build the fraud models.

So how do we solve this challenge? By using self-calibrating analytics to understand both sides of the payment, which helps us make a much more accurate decision on whether a transaction is fraudulent.

We recently released custom account-to-account fraud models that monitor electronic payment behaviors for both the payer and payee accounts. This enables the models to capture typical payees, payment amounts and payment behaviors associated with a payer DDA account, while also incorporating recent payment transaction behaviors to the payee account.

Retail_Banking_Fraud_Management_450-pxa_zol

Self-calibrating analytics determine outlier behaviors for accounts with abnormal payment activity and generate high scores where the payment is likely fraudulent. As I've blogged previously, this type of model relies less on historical data and can adapt in production on the fly as new fraud patterns emerge. Because of this, self-calibrating analytics are ideal for use in situations where there's insufficient data to build traditional neural network fraud models.

In fact, our last DDA model release was designed to protect a client portfolio that has not yet launched in the market place. We worked closely with the client to design an analytic solution that would provide quality fraud detection for the portfolio from day one. And as new FICO clients sign up for DDA models, we'll have more and more consortium data to build better and better fraud detection solutions in the retail banking space.

related posts