There's tremendous emphasis on increasing customer centricity in banking these days, and indeed a big focus of my fraud analytics team is developing “customer-focused” innovations. These analytic models go beyond focusing solely on the characteristics of typical fraud events; they’re also built to improve learning of what typical customer behavior looks like in order to reduce declines of legitimate transactions—what we in the fraud world call false positives.
One such customer-focused fraud innovation is our patented Behavior Sorted List technology. It's designed to capture a customer’s specific spend favorites, such as locations and merchants. As transactions occur, an iterative algorithm updates a rank-ordering of events within the card transaction profile to allow the determination of favorites. In 2013, we're updating all consortium models in FICO® Falcon® Fraud Manager 6 to include these lists, which will provide a material reduction in false positive rates.
Another customer-focused innovation uses a patent-pending streaming version of Latent Dirichlet Allocation (LDA) to summarize the customer’s transactions into customer archetypes based on a global macro view. Using those customer archetypes, the model determines whether new, previously unseen transactional behavior should be viewed as typical or fraudulent. In other words, this analytic method can better distinguish between likely fraud and what's simply new behavior from a legitimate customer.
If you're interested in learning more, I invite you to join us for FICO World 2013 (Miami, April 30-May 3), and specifically my presentation on "Innovations in Fraud Analytics." In that session, I'll dive deeper into these and other customer-focused technologies. I'll also share specifics about the false positive reductions we've seen these methods provide at a given fraud detection rate—a huge bonus for banks trying to find that delicate balance between customer focus and keeping fraud in check.