To fight counterfeit goods, resale platform uses machine learning

Counterfeit goods have been around long before the internet, but with resale platforms flooded with hundreds or even thousands of items a day, it’s difficult to keep all counterfeits out. There are cases. Trained workers oversee the authentication and verification process of these platforms, while machine learning can detect discrepancies at levels that the human eye cannot detect. In many cases, resale platforms employ machine learning to train authentication algorithms to recognize patterns and features that distinguish real from fake. This includes analyzing products for logo placement, trademarks, etc. and flagging any discrepancies.

The RealReal, which specializes in consignment sales of luxury goods, has been using machine learning to retouch images and create product copy since 2018. Over the past few years, however, we have developed additional tools to target counterfeit goods more effectively. About three years ago, The RealReal built a system called “Shield”. “Shield” collects information about the shipper and tells you if the shipper has previously submitted counterfeit goods. But ‘Shield’ doesn’t really focus on the product at hand. To help with this, The RealReal developed another system, Vision, about a year and a half ago. Vision uses camera technology to assist the verification process. The RealReal is currently working on a system that combines the two.

Christopher Brossmann, vp of machine learning at The RealReal, told Modern Retail that machine learning will ultimately “provide decision support.” “It takes a lot of work to get a good signal for authentication, but using this technique can make things a little faster,” he explained. Items that do not meet The RealReal’s standards will be returned to the sender in accordance with the Three Strikes Policy. If The RealReal detects “obvious fraudulent intent”, it will seize the product and hand it over to law enforcement agencies.

StockX only accepts products in pristine condition, and their certification system evaluates every item based on 20-50 touchpoints. This includes everything from checking the packaging to making sure the product is the correct size and color. It also relies heavily on embedded technologies such as chips and IDs that manufacturers themselves embed into handbags and shoes. “Counterfeit products emit different qualities that indicate they are abnormal,” Paul Foley, StockX’s head of brand protection, told Modern Retail.

Many of the counterfeits StockX catches come from people who bought the product elsewhere and didn’t know it was counterfeit, Foley explained. StockX will then return the item to the sender.

StockX has built three new certification centers in Mexico City, Tokyo and Berlin in the last 12 months. Foley said the company is currently building an innovation lab in Detroit to study other technologies that could be useful for authentication systems.

“We are still looking at and weighing different technologies,” he explained. “You have to look and say, ‘Are they worth it?’ How can they scale? We are setting up an innovation lab to explore the product, but it may not necessarily be the vertical where the product is currently being used.”

Meanwhile, eBay acquired solution provider 3PM Shield in February, giving it access to additional technology to prevent the sale of counterfeit goods.

“picturebay employs a comprehensive approach to protecting buyers and sellers through a strong combination of people, policies and technology,” a spokesperson told Modern Retail in an email. “There are also tools supported by artificial intelligence technology, eBay said that in 2022, he will actively block 88 million suspected counterfeit listings and have them reviewed by eBay investigators. In response, he was able to proactively remove 1.3 million items.”

Finding counterfeits has become a bigger problem for the resale market, especially as people show interest in buying second-hand goods. This is especially true in the luxury market, where The RealReal saw a 22% year-over-year increase in active buyers in the first quarter. Additionally, while some marketplaces such as eBay have been around for years, more competitors are entering the space and shoppers are taking their business elsewhere. increase. For example, EBay has recorded a year-over-year decline in active users over the past year. As platforms rush to sell, many are betting on machine learning to prove their products are the best.

However, while machine learning has many advantages, it also presents some challenges. Acquiring and building machine learning can be expensive, and its systems are “as smart as the public data,” said Kassi Socha, a Gartner director and analyst at his firm. told Modern Retail.

Many companies may have years of credible authentication data, he said, but a platform starting from scratch would need to acquire a variety of authentic and counterfeit goods to train its algorithms. It is said that there is Brossman of The RealReal also said that programming these tools requires a certain level of expertise in the first place. “You have to be a certain amount of subject matter expert to do stuff like counterfeit detection,” he said.

Moreover, like any technology, machine learning is not error-free, says Sochi. “Of course, the output of AI-powered machine learning is not always predictable, so the output can change as the model learns,” she explained. “In other words, it is truly a black-box algorithm, so exposure to wrong or insufficient data can compromise its ability to detect counterfeit goods. is a significant risk or challenge.“

There is also the risk that counterfeiters will use this machine learning to get ahead of themselves and create better counterfeits, Sochi said. That’s why it’s especially important for the apparel, footwear and accessories industry to invest in digital identities to help prevent this “second-order effect,” she said.

Some companies are already doing this. For example, Mulberry is introducing digital ID in leather goods, and Prada is rolling out his NFC (Near Field Communication) and RFID (Radio Frequency Identification) tools. Several luxury retailers such as LVMH and Cartier have banded together to form the Aura Blockchain Consortium. This consortium offers, among other things, a certificate that guarantees the authenticity of luxury goods.

Similarly, StockX’s Foley has called on various platforms and brands to work together to curb counterfeit goods.

“They have a lot of good data points,” he said. “We have a lot of great data points, and I think it makes sense to track the counterfeiters together. I think it could be game-changing.”