AI Proves Effective in Stopping eCommerce Returns Fraud

Returns are a cost of doing business for all merchants.

If not handled well, they can present a point of friction, disappointing customers so much that they’re discouraged from shopping with that company again.

Fraudsters? They’re gaming the system, sometimes reporting that products did not arrive, or they did not get what they ordered — and they request refunds.

Doriel Abrahams, head of risk, U.S., at Forter, and Richard Kostick, CEO of 100% PURE, said the same data that is used to fight fraudsters can be used to improve the returns process itself, and in doing so, improve customer loyalty.

The Trends Today

Abrahams said that in the digital age, a popular move is bracketing — where online shoppers buy an item in several sizes and then return the choices that don’t fit. Elsewhere, some studies find that more than half of consumers have confessed to returning items after using them only once or twice.

“We believe that around 15% of returns are actually fraudulent” or represent abuse of the system, he said. The scams are numerous, as individuals return empty boxes to retailers or return boxes laden with bricks. Some scammers even resell the items they have received after being refunded.

For merchants, added Kostick — and from the vantage point of 100% PURE — returns are costly. Staff must be deployed to handle the goods shipped back. For Kostick’s firm, focused on cosmetics, the fact remains that used inventory simply can’t be put back; it must be destroyed. The company shifted to a policy where a consumer can opt to keep the order rather than return it, which has given rise to a different fraud vector, as they also request refunds and keep the cosmetics.

Subsequently, 100% PURE embraced a policy where automated returns have given way to a high-touch process, as the returns continuum demands that individuals reach out to the company’s customer service department. The staffers take what Kostick said are “copious notes” on customers they think are abusing the process. Those same customers are required to ship the products back at their own expense, which cuts down on fraud. Serial abusers are simply cast out of the 100% PURE customer base, as the firm will refuse to do business with them. The company’s returns rate is about 2%.

Data Is a Critical Defense

The 100% PURE approach, with a personal touch in the mix, is tougher to scale when there are millions of customers. Customers can also create new identities online and continue abusing the returns process.

As Abrahams noted, data can make the difference between keeping good, transacting customers in place while weeding out the bad actors. Forter’s platform helps uncover users’ identities as they transact with Forter’s partner firms, mapping out those identities and linking them to specific behaviors using generative artificial intelligence. That same use of analytics and advanced technologies can determine just who’s a good customer and automate returns so that there’s no need to reach out to customer service.

“The key is to teach your AI models and the systems to ‘think’ the ways these people think and ask the right questions at the right time,” he said.

Analytics can uncover if consumers are all coming from the same region, or even a specific device. Kostick added that AI can also be used to improve consumer-facing interactions, as high-touch customer service is delivered at scale.

AI also helps with inventory management, said Abrahams and Kostick, as a high number of returns might indicate that there’s a flaw in a product, the packaging or simply that consumers’ tastes are changing. We may indeed be headed toward a future where personalized products are made on demand, which in turn has a positive ripple effect on supply chain sustainability.

For now, as Abrahams and Kostick told PYMNTS, data sharing will help separate good customers from bad actors. The battle will be a long one.

After all, Abrahams said, “there are people out there whose entire purpose in life is to find loopholes and little cracks” in business processes “and make them into full-blown holes.”

For all PYMNTS retail and AI coverage, subscribe to the daily Retail and AI Newsletters.