At Stitch Fix, the online personal styling service, finding the right thing to wear is as much about fashion as it is about data. The company has hired over 100 data scientists to build a robust set of algorithms that determine everything from the size, silhouette, and style of each item that shows up on a client’s doorstep.
But fashion, as Vogue editor Carine Roitfelt once put it, is never really about the clothes. It’s about the look—the way multiple items come together. Now, Stitch Fix is putting its data scientists to work toward understanding outfits in a more holistic sense, using algorithms.
A new feature, called Shop Your Looks, will suggest items that “go” with a piece of clothing bought through Stitch Fix. The flouncy skirt gets a button-down blouse to balance it out; the wide-leg jeans get a silk kimono and pink flats. The goal is to help clients get more mileage out of each $20 styling fee, and of course, to sell more apparel.
Stitch Fix is just one of several brands racing to shape the future of personalized fashion merchandising. Instagram now sells clothes like a personalized mall. Pinterest offers shoppable pins and recommendations like a personal stylist. Amazon, in its quest to sell you absolutely everything, has also dabbled in fashion. Its latest tool, a program that recommends clothing based on a reference photo, like a “Shazam for clothes,” is meant to re-create serviceable outfits from Amazon’s vast inventory. Amazon calls it a way to “change the way you shop, forever.”
Stitch Fix’s approach is decidedly more complex than just recommending an outfit. The company wants to understand absolutely everything about why we wear what we wear, using data science and math. The company is working to master the peculiarities of sizing, understand the nuances of personal style, and predict what clients will want to wear months before they receive a box in the mail. Now, it’s also using data to demystify one of the most elusive parts of fashion: how individual pieces of clothing relate to one another.
Algorithms of Style
Since launching in 2011, Stitch Fix has tried to offer the kind of fashion advice one would get from a personal shopper, but do it at scale. That means combining, in the company’s words, “an army of stylists with an arsenal of data.” That work goes beyond just using computer models to anticipate the latest trends.
“People think of Stitch Fix as this styling algorithm that picks out clothes for you,” says Eric Colson, Stitch Fix’s chief algorithms officer emeritus. “And sure, we do that. But there are more than 100 data scientists and only six work on that. The rest are doing all kinds of other algorithms.”
For example, the company uses algorithms to determine how to pair stylists with clients and to help buyers predict what will be in style months into the future, so they can better manage inventory. For each client, Stitch Fix uses a complex set of algorithms to decode “latent style”—the types of items a client actually likes, regardless of how that client self-labels their style—and then display those preferences on a map that clusters items by style, with “boho” in one corner and “preppy” in another. It also uses algorithms to understand each client’s “latent size,” since not every medium-size shirt will fit every medium-size client the same way. Over time, these algorithms help to deliver clothes that are likely to best fit the idiosyncrasies of an individual’s body—even in extreme cases, which Colson refers to as the “Michael Phelps problem” in reference to the atypical body shape of the world-champion swimmer.
Source: Need Some Fashion Advice? Just Ask Stitch Fix’s Algorithm
By Arielle Pardes
Techylawyer and its authors do not claim to have written this article, we acknowledge the works of the original author