ADVERTISEMENT

Stitch Fix Uses Netflix-Like Algorithms to Help Customers Find Their Style

Stitch Fix Founder Katrina Lake says even the biggest troves of statistics lack power unless they’re analysed.

Stitch Fix Uses Netflix-Like Algorithms to Help Customers Find Their Style
An attendee uses a touch screen display to try on clothes on a virtual avatar (Photographer: Kiyoshi Ota/Bloomberg)  

(Bloomberg Businessweek) -- It takes guts to tell a fortysomething woman she should wear a romper—the one-piece shorts-and-shirt combination traditionally worn by small children. Still, that’s what Katrina Lake, 36, co-founder and chief executive officer of online retailer and styling service Stitch Fix Inc., did in early June when she pulled together an assortment of clothing for a customer. Besides the romper, Lake added a floral wrap blouse, a tank top, red shorts, and a skirt. “When our algorithm is recommending to me a romper or jumpsuit for a 40- or 50-year-old, I totally trust it,” Lake says, talking about the statistics-driven program she consults before selecting pieces. Human bias, she says, would have counseled against the item.

Removing bias is a large part of the success of Stitch Fix. For a $20 fee, the service mails out clothing to clients based on what it’s come to know about their tastes—not based on their age or ZIP code. Over the last year, Stitch Fix says, 3.1 million people, mostly in the U.S., have used its service. On June 5, amid a terrible earnings season for the apparel sector, the company reported $408.9 million in revenue for its third quarter, which ended in April, a 29% increase from the year-earlier period. It also posted $7 million in net income. Analysts had projected a $2.3 million loss. Stitch Fix expects revenue for the year to rise to $1.58 billion, up from $1.23 billion for 2018.

The recent results helped send shares up to around $30, but before that the company had been a tough sell among investors, who say they’re skeptical it can continue to grow at a solid pace. Late last year, when Stitch Fix reported lower-than-expected revenue, its shares plunged to just under $17. It didn’t help that retail giant Amazon.com Inc. had just invaded Stitch Fix’s space: It brought out Prime Wardrobe, a service that lets Prime customers try clothing for free, and started testing Discover Your Style, a program that makes personalized shopping recommendations.

Stitch Fix Uses Netflix-Like Algorithms to Help Customers Find Their Style

Lake says Stitch Fix’s relatively unusual approach to selling apparel has made it hard to win the confidence of investors, a situation she finds frustrating. “We’ve done what we told the Street we would do,” she says, adding that as investors see she can deliver, she should gain their trust. Stitch Fix opened for business in 2011 and listed on the Nasdaq almost two years ago, with shares priced at $15.

“Their biggest problem is customer retention,” says Sucharita Kodali, a retail analyst at Forrester Research. Stitch Fix can encourage customers to increase the frequency of delivered boxes, increase the number of items in each box, or play with pricing. “They’ve probably tried all of that, and there’s a natural limit.”

Megan Long, 33, a music theorist at Oberlin College in Ohio, tried Stitch Fix about six years ago, when she needed professional clothing. Initially, she was thrilled. The items, including a houndstooth jacket she still wears frequently, fit well and matched the “gently preppy vibe” Long aspired to. In large part thanks to Stitch Fix, she says, she now has a better sense of her style and feels more confident shopping on her own. She stopped using the service about two years ago.

But for every Megan Long, cue a customer like Nancy West, who tried the “transformative” service a year ago and has since received about 10 boxes. “Only once I started using Stitch Fix did I realize how much of a difference it could make just to feel confident that I was dressed nicely,” says the 52-year-old Boston-area freelance writer. Her stylist perfectly understands her request for “age-appropriate, casual, and cute,” and she generally keeps two or three items from every Stitch Fix assortment.

Customers—mainly women, although the company also caters to men and children—receive their packages after filling out a style profile so Stitch Fix can figure out what they like. For the $20 styling fee they get a selection of five pieces in their price range, chosen by a stylist based on the algorithm’s suggestions. They can buy what they like and return the rest. (The $20 fee is applied toward the purchase price, and users get a discount for buying everything in the box.) The more a person uses the service, the better it gets at nailing her look.

If this sounds something like Netflix, that’s by design. The streaming service’s former vice president for data science, Eric Colson, joined Stitch Fix as chief algorithms officer in 2012. (He’s now CAO emeritus.) Just as Netflix Inc. uses one movie as the basis to sometimes recommend a seemingly unrelated one, Stitch Fix looks at detailed attributes to figure out that a customer who typically likes conservatively cut blazers and dresses might go for a black leather motorcycle jacket in her next order. (The unexpected link: She’d been buying more mixes of prints and patterns, indicating an edgier streak.)

Lake already was drawn to this type of expertise as an undergraduate. A data geek, she wrote her honors thesis at Stanford on econometrics and health results, and thus can throw around terms like “multivariate regression” in the same breath as “scalloped edge.” In a consulting job after college that involved working with EBay Inc., she saw how even the biggest troves of statistics lacked power unless they were organized and analyzed the right way. “They had just tons and tons of unstructured data that was really hard to parse out and make useful,” she says. “It was so hard to unlock it.” By the time she started Stitch Fix in her Cambridge apartment while an MBA student at Harvard, she knew she needed to place as much emphasis on data as on style. (A co-founder, Erin Flynn, sued the company in 2012, alleging breach of contract and wrongful termination; the case was settled in 2014, and Flynn is no longer with the company.)

From its beginnings, Stitch Fix gathered information that matters about customers, says Chief Operating Officer Mike Smith, a Walmart.com alumnus. “They give us data about their bra size,” he says. “They give us data about how a pair of pants feels on their thighs.” Tidbits like that help the company refine its take on what will delight customers and keep them coming back and spending more.

Stitch Fix doesn’t define “high-quality customers” by how many times a year they shop, but by how much they spend. The company aims to boost the average annual spend per customer: It’s $467 today, up from $443 six months ago. That focus on customer spend per order was one reason for Stitch Fix’s expansion into the U.K. earlier this year, its first foray outside the U.S., despite the uncertainty caused by Brexit. People there purchase consistently online and don’t expect the same deep discounts U.S. customers look for.

Another challenge to enticing customers to spend more comes from managing the physical supplies of the products the algorithms recommend. “Inventory flows SUCK and no one cares,” wrote one Cleveland-based stylist earlier this year on workplace review site Glassdoor. The company says it’s taking steps to improve the availability of popular clothing.

As to the fear of retail-killer Amazon, Lake has in the past spoken about how, despite its millions of available pieces of clothing, the online giant can’t figure out which garments are best for each customer. That’s a problem Stitch Fix solves, she says. Usually, she leaves the Amazon-bashing to Smith. “We’re not in the cloud, we’re not in logistics, we’re not in movie studios,” he says in a not-so-subtle jab at his competitor’s many business lines. “Delivering against expectations will continue to help with the story.”

To contact the editor responsible for this story: Dimitra Kessenides at dkessenides1@bloomberg.net, Madeleine Lim

©2019 Bloomberg L.P.