Fix’s technical corps. “The tech part of the organization hasn’t kept up with the pace of the rest of the business, so we’ve doubled the engineering team over the last year,” she says.
The Style Shuffle app emerged from internal discussions at the company. The game was initially made available via Facebook Messenger. “We had a waiting list for people to try it out and we realized that we were on to something,” Polinsky says.
So Stitch Fix began developing an app, which it made available to all customers last month. Polinsky says the company has so far received about 200 million ratings from customers playing the game.
Such a jump in new available data is attractive for these e-commerce companies. Customers do fill out detailed questionnaires: Stitch Fix says it asks 85 questions ranging from sizes to parts of the body one would like to camouflage.
But those customers receive, at most, twelve fixes a year, with about five items apiece. That means the ongoing customer feedback results in only about 60 data points. The game, Polinsky says, gives the company dozens of data points from a single encounter.
While playing the Style Shuffle game is optional, Polinsky says her team encourages customers who are not happy with their selections to use it. “It’s a great way to know what’s an upcoming style we should keep our eye on,” she says.
One of the biggest challenges for retailers and brands is finding shoppers who want their products. Polinsky says Stitch Fix has not yet explored ways to share the data they are receiving from the game with brands. They currently provide other data based on customer feedback such as the fit in the shoulder of a product is too narrow or a specific size is running too large. “There are adjustments they can make in the next manufacturing batch,” she says.
While Stitch Fix is understandably focused on strengthening its machine-learning capabilities, I asked Polinsky whether she sees a need to integrate them with other important e-commerce technologies, such as augmented reality or visual search.
“They’re not core to the matching technology that we’re doing today, but I could see that there could be opportunity there for understanding new trends and styles and discovering look-alikes for customers,” she says.
