Adidas improves online outfit recommendations with the use of machine learning

"Before the solution, Adidas merchants would have to manually put together outfits for its online “Complete the Look” feature — a process that took 27 steps and 20 minutes to finish, and resulted in fewer than 10 percent of products appearing in the feature. And since the customer average order value (AOV) significantly increased when they were shown a recommended outfit, Adidas wanted to find a way to make the process easier."

"Adidas is partnering with artificial intelligence (AI) platform provider Findmine to offer complete product recommendations on the retailer’s site. The German sportswear company can automatically generate complete recommended outfits as a customer browses the website looking for individual products. Enter Findmine, with an AI solution that was initially tested out during a six-week period that showed half the traffic to the Adidas site manually-programmed outfits, while the remaining half were offered outfits generated by Findmine. The pilot proved that both merchants and customers were unable to distinguish the difference between the AI-based outfits and the manually-created ensembles. “Findmine has helped us reduce the amount of manual work and has helped us ensure that our newest products have cross-selling from day 1, improving conversion, average order value, and customer satisfaction better than any other solution we’ve tried,” said Bryan Klavitter, senior director, Adidas consumer experience."

"At my company Findmine, based at NYU Tandon’s Data Future Labs, we teach our system how to learn why an outfit or combination of furniture works, so it can make its own decisions. It allows us to take our customers’ entire product catalogue and build outfits on the fly or keep up with changing inventory — even changing seasons and trends! — because it’s figured out what works and what doesn’t, constantly readjusting to information." (venturebeat)

Outfit pairings