Few American startups yield more insight about the future of business than Stitch Fix. To call this San Francisco-based company an online fashion retailer doesn’t begin to give a sense of its uniqueness. It grew to $730 million in 2016 revenues from a start in 2011 when founder (now CEO) Katrina Lake was at Harvard Business School. It sells women, and more recently men, fashions they didn’t know they wanted. But it achieved these enviable results by measuring and optimizing everything. It so suffuses itself in data analytics and artificial intelligence that to hear how it operates feels like receiving brain waves from the future.
But this message is particularly welcome, because of the way Stitch Fix succeeds—by blending machine intelligence with the real human intelligence of 3400 fashion stylists, most of whom work part time from home, interacting with customers. The company’s growing employee base suggests that a future world infused with AI may not decimate the workforce after all.
“It really is the convergence of man and machine,” says Chief Algorithms Officer Eric Colson, a six-year Netflix veteran before joining Stitch Fix in 2012. “We’ve stumbled onto being able to combine machine learning and AI with human judgement, to create a product much better than people could do on their own.”
When a customer first starts “shopping” at Stitch Fix, all she does is fill out a lengthy questionnaire. If you say you wear medium-sized blouses, it asks whether they typically fit you loosely or tight. Does your office require business attire or is it casual? Do you take wardrobe risks? Which of this list of 15 colors would you wear? Please answer definitely, never, or maybe. Do you wear your jeans skinny, straight, or both? Many questions don’t require a yes or no, recognizing that our choices can be mutable. Then the company feeds all this, along with social media profiles, Pinterest boards of styles you’ve “pinned,” and plenty more, into its computers.
The customer agrees to receive clothing sight unseen. “There’s no customer choosing,” explains Colson. These shoppers-who-don’t-shop pay a $20 “styling fee” to receive each periodic box of clothes, but can send anything back for free. The company pays shipping both ways if there is no sale. “We face severe penalties if we get it wrong,” Colson continues.
A box of clothes—called a “Fix shipment”—is assembled by a sort of teamwork between computers and the stylist. A custom data-crunching “styling algorithm” selects a group of items it thinks the customer would like. But crucially, that selection is shown not to the customer but to the stylist, who is generally matched to the customer, again, with an algorithm. The stylist streamlines the computer’s choices to select the ultimate assortment for the customer.
“It turns out there are things humans can do much better, and are likely to remain better at for a long time,” says Colson—”things like curation, the ability to see things as a cohesive set, and to improvise. Not to mention being able to relate to other human beings.”
And the stylists seem to love their jobs. Stitch Fix surveys their job satisfaction twice a month. Whereas in most companies part-timers are the least satisfied employees, at this one they are the happiest.
So employment is growing in this smart, modern company because computers are augmenting what people do. That is a big, fat, positive data point for a world where software, as they say, is eating the world.
David Kirkpatrick is chief techonomist.