A consumer has spotted the perfect outfit for an upcoming night on the town. She snaps a photo of it and asks the person wearing it where she bought it. All that remains now is finding it on the store’s website based on the photo.
But that’s the problem: Most electronic commerce search engines work based on keywords, not images. Typing a description of the outfit into a search bar might lead to many frustrating hours of hunting, quite possibly in vain — even though this shopper knows what she’s looking for and who sells it.
Her best bet is probably to take the photo to a retail outlet and hope a sales clerk recognizes it. So much for digital transformation in retail.
“Fashion is all about the visual,” said John Xiao, vice president of technology for the department store chain Nordstrom. “If the customer has a product in front of them, the best way to help them find it in your catalog is actually visual search.”
Google Offers Four New Products for Retailers
It’s a problem for retailers that Google says it is trying to solve with Vision Product Search, one of three new products it made widely available to retailers this week at Google Cloud Next ’19 in San Francisco. The three products, plus a fourth launched in beta mode, use advanced AI and machine learning to help retailers deliver new customer experiences, make better use of data and drive revenue.
“We’re going through a time of transformation,” said Mark Regan, a Google Cloud product manager. “It’s a difficult time in retail, for sure. Customers expect a personalized, engaging shopping experience. But it’s also an exciting time.”
Each of the new products will be familiar to most retailers. Visual search, for example, has been around for several years, and a number of retailers have experimented with it. However, Google says its versions are more effective than what’s currently available.
Nordstrom, an early adopter of Vision Product Search, found that it did indeed do a better job of locating the correct product than its legacy system, according to Xiao, which was accurate about 80 percent of the time.
“That actually isn’t good enough,” he said. “This new platform cut our services cost and increased our accuracy to over 95 percent on day one.”
Better Consumer Recommendations with Cloud Tech
The other products Google announced this week are Recommendations AI, which Google says delivers more thoughtful suggestions to consumers based on their search and buying choices than other recommendation engines; AutoML Tables, which uses machine learning to help retailers do more with their structured data; and, in beta, Contact Center AI, for natural language-driven call automation.
Kathy de Paolo, vice president of engineering for The Walt Disney Company, said its early adoption of Recommendations AI had a noticeable effect on shopDisney, its e-commerce arm. For example, its previous recommendation engine might show a consumer looking for sleep masks a mix of appropriate and inappropriate products, such as Halloween costume masks. For a shopper who’d purchased a red dress, the engine would recommend other red dresses.
“They probably don’t want another red dress,” de Paolo said. “Google figured out the buyer is more likely to buy something to go with the dress, like shoes or accessories.”
Since deploying Recommendations AI, shopDisney has seen increased order value, units per transaction and overall revenue, she said.
Keep this page bookmarked for complete coverage of Google Cloud Next '19.