Banks are engaged in a never-ending battle to stay a step ahead of the fraudsters, who are clever and nimble enough to constantly pivot to exploit weaknesses in the financial system. Business accounts have long been a popular target since, as bank robber Willie Sutton supposedly said, “That’s where the money is.” However, the tactic has recently morphed to involve larger volumes of small-dollar transactions, which are harder to detect.
We’d like to share some details regarding the success a fast-growing $20 billion US bank realized by installing ARGO’s OASIS cross-channel fraud detection platform. With rapidly expanding volumes and a wide array of existing systems, the FI needed a better fraud detection capability without additional manual effort, and without depending on legacy solution providers for implementation.
Artificial Intelligence Brings Meaningful Intelligence
The key to the exercise was a reduction in false positives. Flagging a lower percentage of transactions for human intervention underpins multiple benefits. It frees up bandwidth to accommodate growing volumes, improves customer experience and enables risk analysts to focus their expertise on the cases that really matter. Of course, this is only valuable if no additional cases of actual fraud (false negatives) are allowed to slip through in the process.
The bank’s automated solution needed to validate check signatures (including dual signature where required), examine large dollar amounts and isolate other suspicious items for a variety of factors that naturally evolve over time as new schemes are discovered, through machine learning as well as manually.
Applying Human Expertise Where It Matters
The results were impressive across all dimensions, beginning with an implementation time of a mere eight weeks with minimal disruption to existing operations. The share of transactions routed for manual review has fallen to 0.7%, well below the industry standard of 1-2%. This has allowed the FI not only to screen more of its transactions through the enhanced AI engine (over 4.8 million items in a recent fiscal quarter), but also for its professional fraud screeners to work more promptly and effectively in addressing the roughly 350 questionable items referred to them each day.
The FI’s false positive to false negative ratio now stands at an industry leading 45:1, as compared to a more common metric of 200:1. The most relevant metric of all- the FI’s rate of check fraud- has also declined, with the solution being proven effective against both first- and third-party fraud. You’ll find more details on this case study here.
This battle is never permanently won. Savvy fraudsters will eventually decide to redirect its efforts toward an FI with less sophisticated fraud detection, or more likely they’ll concoct new schemes to probe for additional system vulnerabilities. The good news is that state-of-the art detection systems apply continuous learning through behavior pattern recognition and updated decisioning/scoring to stay a step of the bad guys. The first step, however, is for a bank to adopt such a system.
For more information on how to prevent check fraud, download our solution brief.
