UPS deploys AI to spot fakes as holiday returns surge

sign up here.

These retailers are among the many US businesses that have fallen victim to return fraud. Return fraud is a theft scheme in which a customer requests a refund for a retail item and then returns another item of lesser value, such as a cheap counterfeit product that cannot be resold.

Happy Returns' AI tool, called Return Vision, helps spot fraudulent returns by flagging suspicious packages, analyzing their contents and sending them for a final human audit who can verify fraud and withhold refunds, Sobie said.

Happy Returns specializes in unboxed, unlabeled returns. Shoppers bring unwanted items to one of about 8,000 so-called “return bars” inside Ulta Beauty, Staples and UPS stores. There, employees scan, bag, and label items, which are then packed into large boxes and sent to a processing hub each day, saving retailers time and money. Shoppers and scammers love this service because it's easy and often provides quick refunds.

Testing of Return Vision began in early November, and additional retailers will begin trialling the tool later this month, when returns surge during the holiday season, Sobie said.

Jim Green, director of logistics and fulfillment at Everlane, which sells cashmere sweaters and other clothing primarily online, said the initiative is aimed at tackling issues that complicate costs for retailers.

Returns are already hurting profits because of the costs of shipping packages, preparing products for resale and restocking shelves, Green said, adding that 85% of Everlane's online returns in the U.S. are processed through Happy Returns' in-person return and consolidation network.

“It's a double whammy if the real thing doesn't come back. It costs us alone hundreds of thousands of dollars a year,” he said.

Representatives for Under Armor declined to comment, and Revolve did not respond.

According to the report, about 9% of returns are fraudulent.

Happy Returns executives say their AI program only helps identify when incorrect items are returned. It does not address other issues such as “wardrobe” when customers return worn or damaged items.

structure

Happy Returns said its AI fraud detection tools start working the moment a shopper initiates a return online.

Flag returns that are initiated before or immediately after the item is delivered, entries from shoppers with multiple linked email addresses, returns from individuals who have previously engaged in suspicious activity, and more.

Employees at return locations have access to photos of items that need to be returned when they scan unwanted items into the Happy Returns system. You can reject obvious discrepancies.

“Humans may not always be able to pick up on the small differences between the returned item and the purchased item,” CEO Sobhi said.

When returns arrive at Happy Return hubs in California, Pennsylvania, and Mississippi, human auditors open flagged packages.

The company says they take a photo, which is fed back into an AI tool and compared to images and other information about the item expected to be returned. A human team then reviews the AI ​​evaluation and makes the final decision.

“If you're trying to return a pair of $300 boots and you show up with a pair of dirty old sneakers, you're going to get caught right away. What Return Vision does is add an extra layer of protection for some of the less obvious cases,” Green said.

Less than 1% of returns in the Happy Returns network are flagged as likely to be fraudulent by the tool, and approximately 10% of flagged items are ultimately confirmed as fraudulent, the company said. The average value of each scam is approximately $261.

Juan Hernandez Campos, Happy Returns' chief operating officer, said the tool is becoming increasingly important as scammers become more sophisticated.

“Bad actors adapt. We need to adapt, too,” he said.

Our standards: Thomson Reuters Trust Principles.opens a new tab