Consumers returned nearly $900 billion worth of merchandise in 2024, and while retailers have long treated returns as a routine cost of doing business, a surge in return fraud is pushing companies to adopt artificial intelligence tools to protect their bottom line. The average return rate last year was about 17%, according to industry data, and retailers report that fraudsters are exploiting flexible policies by sending back cheaper or entirely different items than those originally purchased. Against that backdrop, logistics firm Happy Returns, which handles millions of returns for major apparel brands, says return fraud has climbed to record levels, with as many as one in nine returns being fraudulent.
Happy Returns CEO David Sobie said about 1% of returns dropped off at the company’s locations are flagged as high risk. Of those flagged returns, roughly 13.5% are confirmed to be fraudulent, underscoring how a small but costly share of returns is driving losses. Sobie noted that while streamlined return systems are convenient for shoppers, they also lower barriers for bad actors who abuse lenient processes. To counter this, Happy Returns has rolled out an artificial intelligence system called Return Vision that assigns a fraud risk score to each return, evaluating customer behavior and product details in a way Sobie likens to how a credit score assesses financial risk, and then red-flagging certain customers for extra scrutiny.
Return Vision’s risk engine looks at signals such as whether there has ever been fraud associated with a shopper’s email or physical address, which Sobie said can trigger a high-risk score, and also examines the time between delivery and when a return is initiated, since a return started suspiciously quickly suggests the buyer may never have opened the package. Items that score as high risk are routed to a secondary screening step where artificial intelligence image analysis compares the returned product to the original. Senior product manager Kayla Hunter demonstrated how, in one case, a pair of designer jeans that looked identical to the listing online was flagged because artificial intelligence detected subtle differences in the waistline, revealing the item was a cheaper duplicate rather than the $298 product that had been purchased. In another example, artificial intelligence determined that an orange sweater matched a $36.99 version instead of the $200 sweater the customer claimed to send back.
When fraud is confirmed, Happy Returns withholds refunds, returns the item to the retailer and flags the associated customer account to prevent further abuse. Hunter said the company feeds this information back into upstream systems so that a shopper’s email and physical address will be recognized and flagged if they attempt similar returns in the future. Retailers warn that ordinary consumers may feel the impact of widespread fraud as companies respond by tightening return windows, imposing new rules and charging more fees on returns to offset mounting losses, even as they continue to lean on artificial intelligence to keep legitimate returns as seamless as possible while filtering out bad actors.
