Returns can be risky business. Here’s how AI can help

Last month, a man was surprised to find pasta instead of Legos in his supposedly new Star Wars Lego kit. Thieves had purchased the kit online, replaced the colorful blocks with dried noodles that mimicked the weight and sound of actual Legos, and returned the doctored box. The unwitting retailer then resold the kit to the equally unsuspecting Star Wars fan.  

This type of return fraud isn’t uncommon, says John Drechny, chief executive officer of the Merchant Advisory Group, a global organization representing more than 200 of the world’s leading merchants across numerous industries. It’s called “bricking” — replacing high-value merchandise with something less expensive, then returning it for a refund.

Merchants of all sizes have been fighting return fraud nearly as long as there have been buyers and sellers. Fake receipts, empty boxes and “wardrobing” (purchasing clothes and returning them, post-wear but with tags in place) are just a few of the ways consumers try to abuse the system. Return fraud in all its iterations is growing at an alarming clip, costing merchants billions of dollars every year.

“It’s not just the fraud alone,” Drechny says. “There’s tremendous effort and cost in protecting against it.” For example, in 2023 TJ Maxx and its affiliated stores ultimately lost around $100,000 when two people used fake receipts to return luggage and handbags across 23 states. 

What sellers need is help reducing the time and costs associated with protecting against fraud, investigating and fighting it. The problem has always been that merchants, whether big box stores or mom-and-pop shops, usually don’t know if a specific return is fraudulent until it’s too late.

Now a new generation of AI-driven tools is starting to change that, says Seckin Yilgoren, Mastercard’s senior vice president for Security Solutions, North America. AI can turn patterns buried in vast volumes of transaction activity into real-time signals at the point of the return, he says, flagging a criminal customer (or just a confused one) and halting the dispute process before it turns into a costly chargeback. 

There are several reasons for the increase in return abuse, starting with the rise of online shopping and the way merchants strive to make it convenient for consumers to buy — and return — from home. That frictionless return process, especially those not requiring a face-to-face interaction, is a big contributor. Consumers can claim a package didn’t arrive, the box was empty or the product was damaged in transit.

And most online returns can be shipped, no in-store visit necessary. Return fraud requires no particular skill set or specialized equipment. There are even social media influencers who promote return abuse as a fun, “victimless” crime. But that’s simply not true, Drechny says.

“Somebody is paying, and that person has to absorb the cost in some way,” he says. “It could be anything from the retailer may have to let people go to they may even go out of business.”

That’s because the costs are significant. In 2024, Mastercard processed $125 billion in returns — more than $17 billion of which was estimated to be fraudulent, Yilgoren says.

A fast, easy consumer experience is the end game for most merchants, but trying to deal with return fraud is making that harder to achieve. Retailers are adding shipping charges and restocking fees and starting to require more paperwork for returns. As a result, legitimate consumers might not return to the merchant if they have to jump through excessive hoops just to return an item.

After consulting with MAG about the specific needs of its members, as well as talking to small and medium-sized businesses as well as consumer card issuers, Mastercard this month launched Return Risk Intelligence, which is designed to help stop return fraud before it happens.

Return Risk Intelligence uses AI to analyze current and historic anonymized Mastercard transactions, specifically returns, refunds and dispute records. It identifies patterns and behaviors that correlate with reported fraud or chargeback events, using this intelligence to generate scores that help merchants predict which returns are more likely to be fraudulent. That gives merchants the ability to make better decisions in real time about accepting a return. It’s part of Digital Enablement, strengthening Mastercard’s suite of capabilities addressing fraud by enabling intervention at the merchant level.

“It's not about alleging a particular customer is committing fraud, it’s just additional insights and data points to help a merchant mitigate risk,” Yilgoren says. “We see a massive volume of transactions, and we use that data pool to find anomalies so we can identify risk and flag behaviors.”

That might look like a single credit card number involved in dozens of returns from numerous sellers, a credit card number that’s been issued thousands of dollars in refunds over a short period, or one that seeks returns frequently on high-priced items, among other signals. Merchants could respond to flags by not immediately issuing a refund or exchange. That way, they can take a day or two to research the requested return and decide if it’s valid.  

While Return Risk Intelligence might not prevent a would-be thief from physically packing a Lego carton with pasta, it just might raise a red flag when that crunchy-sounding cardboard box lands in the returns bin and the sender’s credit history is automatically reviewed.

“Mastercard has a better view overall of the ecosystem, where a merchant can only see what’s happening at their store,” Drechny says. “It’s a much broader perspective to make decisions based on.”