The FICO analytics team: David Griegel, Todd Smith, Yuting Jia, Alexei Betin, Jun Zhang and Scott Zoldi. Credit: FICO
Nearly half of all credit card fraud in the world occurs in the U.S., according to a report by Barclays.
To help stop the growth of sophisticated fraud, FICO, best known for providing credit scores, developed a system that combines payment authorization data from a global consortium of its banking clients.
Known as Falcon Fraud Manager, the system uses streaming analytics and a central network scoring engine to create a real-time transaction monitoring and decision platform.
With the data and technology, FICO can develop fraud models to help its clients differentiate between legitimate and fraudulent transactions, enabling them to make real-time decisions to prevent loss.
Having clients contribute data for the project raised several challenges, says Scott Zoldi, vice president of analytic science in FICO's product technology organization. The volume of data was imposing, and FICO had to find a way to ensure the quality and consistency of the data across clients. What's more, FICO receives and processes billions of data records every month, so it had to develop processes that would systematically inventory, cleanse and prepare the data for data scientists to quickly access online.
Today, FICO protects over two-thirds of the world's payment card transactions from fraud with Falcon Fraud Manager. "Very often, we find clients see more than a 50% reduction in fraud losses on an individual basis when they go from an in-house system to Falcon," Zoldi says.
Zoldi advises companies that are considering streaming analytics to place as much emphasis on hiring talented people as they do on deploying good technology. "I often hire Ph.D.s on my team, but it takes a lot more coordination in understanding the business problems and the IT constraints," he says. "There's business value to minimizing fraud loss, but not at the expense of adverse customer impact."
Sign up for Computerworld eNewsletters.