For the latest AI Innovators interview, we asked Eusong Huynh, Vice President, Customer Success, about acquirer fraud. He explained the shifting ecosystem and why early detection is more important than ever before.
With card-not-present transactions growing twice as quickly as card-present transactions globally, the risks and the ecosystem are shifting from the issuers to the acquirers and their merchants. Acquirers are now taking on more liability, requiring robust acquirer fraud solutions that detect and shutdown fraudulent merchants quickly and efficiently. Payments fraud results in massive chargebacks, up to 20 percent annually. Additionally, Ethoca reports, 52 percent of orders thought to be fraud turn out to be good orders that result in lost business for merchants.
Acquirers have traditionally used rules-based systems to deal with fraud. Now with growing liability risk, they are developing two distinct problems: more alerts than can be investigated and growing staffing needs to handle those alerts.
With such a high volume of alerts, acquirers’ fraud management teams can only comb through volumes of data looking for the most obvious anomalies while missing others. Regardless of the number of resources on staff, the work is overwhelming.
Reducing acquirer fraud alerts and increasing accuracy
Using an AI-based solution to identify fraud provides more accurate results while adaptively learning new fraudulent behaviors. As behavior changes, AI uses machine learning to adapt the profiles of the acquirer’s merchants, their transaction types, volume, and more. Alerts are early and accurate, flagging unprecedented or unusual behavior for investigation. This heightened accuracy brings an immediate ROI, not only in terms of recovering fraudulently acquired money, but also through real-time data uploads, continuous behavior updates, and more focused investigation staff.
Early detection needs speedy processing and output
“It doesn’t matter if you can build a model in a lab that can run a batch that takes hours to compute but gives you great results. You need to put it in a production environment, run it against live data and get your results quickly enough for acquirers to take action on,” said Eusong Huynh, Vice President, Customer Success at Brighterion. “Having a model that can turn out results much more quickly gives the acquirer more time to investigate other alerts that are coming out and truly determine if they are fraudulent or if they’re genuine.”
Assessing risk for acquirer fraud
One thing we’ve learned is that the more data we have, the better the result, Eusong notes. “When you’re following the journey from merchant onboarding to merchant risks to merchant fraud and AML regulatory screening, you get a full 360-degree view on what that merchant is doing, how they’ve progressed from the time they’re onboarded into your system, to how they’re performing today.”
For example, a long-term merchant who is facing financial hardship may try to process several fraudulent transactions or launder money. An AI-based solution quickly identifies that unusual behavior based on its stored data of that user’s expected transactions.
“You have to pair that with a solution that provides information on that merchant in general. At Brighterion, we offer a case management solution that marries in demographic data about the merchant,” Eusong explained. “That allows a single source of information for an acquirer to monitor merchants from a transactional behavioral standpoint, but also pull in this demographic data in order to be able to monitor things like changes in ownership, what they’re selling and average ticket values.”
What if an acquirer doesn’t have in-house AI expertise?
One roadblock Brighterion has discovered in its research is that some organizations are nervous about implementing AI because they don’t have in-house expertise in AI. It’s not necessary, according to Eusong. He said the technology was designed to be powerful and easy to implement.
“We have a program called AI Express that really facilitates that. We’re able to take an acquirer’s data, develop a model in house, and provide a production-ready model within six to eight weeks. Then we take that model and move to a production setting where they can then begin actioning upon the results of the model,” Eusong said.
He explained that Brighterion’s customer support and solutions engineers help new customers step by step. They help identify the necessary data, build the working model, and then provide hands-on support to customers to get their new solution up and running.
“We have institutional knowledge with implementing solutions for large acquirers. We bring that information to help new customers find the right data elements,” said Eusong.
Watch the interview with Eusong Hyuhn on our website as he discusses other ways Brighterion’s AI can help acquirers manage their risk and prevent acquirer fraud.