In this atmosphere of rising fraud risk, banks and merchants often inadvertently block legitimate transactions, frustrating consumers who may shove the affected payment card to the bottom of their wallet or eschew that merchant altogether. The
Two years ago, San Francisco-based Sift began work on a new approach to making transaction-based risk decisions with an AI-powered engine that takes over the job of tightening and loosening fraud filters typically handled by in-house fraud experts, according to a Thursday announcement.
The result is RiskWatch, which leverages AI to automatically account for changes in consumer behavior in real time by limiting more transactions when risks surge, as Sift detects broader fraud attacks happening in real time or during seasonal fraud surges. The tool also automatically relaxes the risk criteria when Sift detects fewer threats.
"Typically fraud experts are making these decisions on a daily or weekly basis about what percentage of fraud transactions to block using their best available information, such as a static risk score, plus a lot of guesswork," said Armen Najarian, chief marketing officer at Sift.
RiskWatch is a "set it and forget it" tool that automatically adjusts a company's fraud filters around the clock based on real-time conditions, so that fraud experts can focus on other tasks, including analyzing specific or more complex fraud cases that require manual review, he said.
Over two years of piloting the technology with merchants including a gaming firm and a cryptocurrency provider, RiskWatch significantly improved the accuracy of users' fraud filters, including reducing the number of good transactions that were blocked, in rigorous comparisons with and without the tool, Najarian said.
"Typically, a fraud analyst at an organization will use available data to set a rate of how many transactions they'll accept, but with the number of fraud variables constantly changing, it's becoming very difficult to account for the surges in fraud risk we see in the market, given different levels of attacks going on and seasonal waves of fraud," he said.
Sift's tool applies AI to real-time data plus data from more than 1 trillion transactions it sees annually across its global network to provide an ongoing, dynamic risk score for an organization's transactions at a speed not previously available, according to Najarian, who noted that Sift has applied for a patent on RiskWatch.
The tool was built using traditional AI in new ways that have evolved along with expanded reach to consider more global fraud data, he said.
"The result is a much more accurate way to apply fraud filters, optimizing the acceptance rate for merchants and banks, in turn, whose good card transactions are less likely to be turned down due to overzealous fraud filters in an increasingly risky environment," Najarian said.
