The Need For Speed
2 comment(s)
In an earlier post, I wrote about the importance of having access to large amounts of contextual data in the fight against fraud (The One with the Most Data Wins). In addition to lots of data, detecting and stopping fraud relies on timely action by financial institutions. Waiting too long enables fraudsters to drain funds, defraud valued customers, and damage the reputations of banks and credit unions.
But how fast does fraud detection need to happen? As with many complex issues, the real answer is, "It depends." Fraudsters perpetrate their crimes in a variety of ways. Some fraud attacks play out over time – days, weeks, or months – and can only be detected through analysis of complex behavioral patterns over this prolonged period. Detecting these patterns requires a batch processing approach that can effectively aggregate a daily volume of hundreds of millions of transactions and application events from multiple sources, find and track the fraud patterns as they emerge, and alert fraud analysts at the appropriate time in the fraud cycle. Other fraud attacks emerge quickly, are fast moving, transactional, and often short-lived. Preventing these attacks requires rapid, real-time detection and transaction blocking supported by analytics that can respond in milliseconds to detect fraud among thousands of concurrent transactions.
The true measure of a successful fraud prevention system is its ability to stop fraud before the point of "value transfer". However, as is highlighted in the table above, value transfer occurs at different points on the timeline for various payment instruments and processes. An effective fraud detection solution must take into account these varying points of value transfer and adapt accordingly. After all, detection after value transfer results in a situation that no bank or credit union wants to face—a recovery effort. In future posts, we'll dive deeper into the concept of value transfer.