SAN FRANCISCO, United States — When Melora Simon goes shopping for clothes at the mall, she has no trouble finding what she likes. “I use labels and brands to help signpost,” says the 37-year-old mother of two, who is currently employed by Stanford University’s School of Medicine. “I know I like the look of J. Crew, the aesthetic of Theory and Diane von Furstenberg. I kind of go there and end up spending more money than I need to.”
But after Simon had her second child, her trips to the mall became less frequent. That’s when she noticed a friend’s Facebook post about Stitch Fix, a San Francisco-based personal styling service that uses a mix of data, algorithms and human touch to send customers boxes of five handpicked fashion items. There is a $20 styling fee attached to every box — or “fix,” as they are called by the company — which can be applied to a purchase. Customers have three days to decide whether or not they want to keep any of the items — and must send back the items they don’t want. Unlike a traditional subscription box model, Stitch Fix users can choose whether or not they want boxes to arrive at fixed intervals, or to request one box at a time.
For Simon, the concept held real appeal. “At that point in my life, my second child was teeny tiny. I had no time to shop and had been in a combination of maternity and breast-feeding clothes for five years,” she laughs. “I was super excited to reinvigorate my wardrobe, but didn’t have a lot of time to go shopping. I loved this idea of getting the chance to tell somebody about myself and my preferences and getting a box shipped to me, with no commitment required, for me to try on in my own home.”
The first box was a “total miss,” Simon says. “There was nothing I liked, but I didn’t want to lose my styling fee, so I gave the scarf to my nanny.” However, when she told Stitch Fix she was dissatisfied, the team persuaded her to give the service a second chance, offering to waive the styling fee on her next “fix.” Like almost half of Stitch Fix’s active clients, she shared her Pinterest profile with her stylist and ended up keeping three out of five of items from the second box, including a polka-dotted skirt, a black-and-white striped fit-and-flare dress and a chambray blazer.
We have so much control. We understand so much about our product and our customers. When we say that something is going to have a 50 percent chance of being kept, it is [actually] a 50 percent chance.
More than a year later, “I still wear those things all of the time,” she says. Since then, the Bay Area resident has used the service sporadically, around holidays or pegged to vacation trips, with varying results. Sometimes she sends it all back; other times she keeps almost everything. But she continues to use the service because it’s enjoyable. “The prices are low enough that I don’t have to think about it,” she says. “It’s a source of newness and fun.”
Indeed, for many Stitch Fix users, the service is often as much about entertainment as it is about buying clothes. But novelty is fleeting. At Stitch Fix, which was founded by Harvard Business School graduate Katrina Lake in 2011, “relevancy” is the key, says chief algorithms officer Eric Colson, who spent three years as vice president of data science and engineering at Netflix before joining Stitch Fix full time in 2012. He now leads a team of 77 data scientists, 48 of whom are PhDs. The algorithm developers come from a multitude of disciplines, including statistics, mathematics, computational neuroscience, systems biology, cognitive psychology and physics. Five of them are astrophysicists.
Getting things “just right” — à la Goldilocks — was one of Lake’s main motivations for starting the company. While many brands and retailers believe customers don’t know what they want until they see it, Stitch Fix uses a mix of direct feedback and carefully tracked consumer behaviour to elicit purchases of items that are a safe bet. The data — “millions and millions” of points in total — is culled from questionnaire responses, email correspondence, reviews, social media posts, purchasing behaviour and more. “We do learn over time that there is a difference between what their stated preferences are and what their true preferences are,” Colson says. “People say things like, ‘Oh gosh, I’m really preppy.’ But then you send them preppy things and it turns out that they have mistaken preppy for what we call casual.”
“I had observed that there was a lot of innovation happening around getting something to someone in the cheapest and fastest way,” Lake says. “But there was no way to figure out on the internet what’s going to be the best for you and your body, which is really what matters when you’re buying clothes.”
From the beginning, Lake knew data would have to play an outsize role. “At the very simplest level, I knew knowing that a certain pair of jeans were good for short people or a certain black dress was good for those who are curvy — simple parameters — offered a real opportunity,” she continues. “Even when I was testing this in my apartment and using Excel, not having any real algorithm, I captured the attributes of clothes, the attributes of people.”
In the complex market for fashion, many have questioned whether Netflix-style recommendation algorithms are the answer. But Stitch Fix posits that its blend of data and human empathy is fundamentally different. “The big advantage we have is a symbiotic relationship with the customer,” Colson says. “We understand that it’s time consuming and difficult to pick out clothes. We’ll take care of that, but in exchange, [the customers] realise that they are going to have to provide some information about themselves. That’s something very different from Netflix.”
“With Netflix,” he continued, “the customer can pick out her own TV shows and movies, so she’s less willing to provide information. For clothes, you really have to do your research. That’s where people want help.”
Customised shopping, where each customer has a differentiated experience, is something e-commerce sites have been aiming to deliver for years. But in the latest installment of Mary Meeker’s famed annual Internet Trends report, the former Wall Street technology analyst, now a general partner at venture capital firm Kleiner Perkins Caufield and Byers, cites Stitch Fix’s approach as setting “the new normal” in retail.
Increasingly, retailers — including Stitch Fix and, of course, Amazon — are leveraging data to launch private label brands that fill gaps in product assortment and implementing “new distribution models enabling direct-to-consumer commerce in the home,” as Meeker points out in her report. Truly customising the experience adds another layer.
Consider a tough-to-fit category like denim. “Finding jeans that fit is a very transformational experience in general,” Lake says. “One of the things we’re able to do is to match people on inseam. Over time, we’re able to understand which fabric and fabrications are more likely to work for different people.”
Tackling universal problems like the search for the perfect-fitting pair of jeans may explain Stitch Fix’s claim that its customer base is virtually demographic-agnostic; it has clients in every corner of the United States and plenty in the middle. The range of product suggests that whoever the consumer is, she is not particularly label-driven. The average price point for a Stitch Fix item is $55, but starts at $20 to $30 for items like costume jewellery and tank tops, and can reach $200 to $300 for dresses and coats. Instead, the company has concentrated on making sure the product mix is correct, developing six in-house labels to fill gaps in the assortment. Today, the buying team works with over 200 brands. And while there is no formal mandate to stylists to favour items from the company’s six private label brands over others, there is a high chance one of these products will land in one’s box. This also ensures Stitch Fix can offer items at a price point that better rivals the likes of fast fashion. For instance, the company created a year-round cardigan to account for America’s perennially chilly air-conditioned offices that costs $48.
Perhaps the best example of Stitch Fix’s data obsession are its three “frankenstyle” pieces, as Colson calls them. The three blouses were developed by genetic algorithms, where different elements from different styles were pieced together. “Frankenstein created his monster by combining different body parts from different people. In the same way, we borrowed a sleeve from one style, a silhouette from another and a colour and a pattern from another still,” Colson explains. “The data suggested that each of those things on their own were very successful with a particular audience. Combining them had signalled that they would be even better together.”
But how many of Stitch Fix’s customers are making a habit out of the service? The company says that 70 percent of clients return for a second “fix” within 90 days and 39 percent spend over half of their apparel wallet share with the service. The internal goal is to get the client to keep one to two items in every box, on average. “To my knowledge, there is no other business out there where 100 percent of their revenue is based on recommendations,” Lake says. “We have so much control. We understand so much about our product and our customers. When we say that something is going to have a 50 percent chance of being kept, it is [actually] a 50 percent chance. And that’s really about the accuracy and relevancy of the data.”
Stitch Fix declined to disclose revenue figures or number of active users, and those close to the company say that the figure most often tossed around — $250 million in sales in 2015 — is incorrect. While technology industry trade site Recode reported in 2014 that the company raised a Series C round of funding amounting to $30 million, the company will not confirm this, acknowledging only that it has raised $16.75 million in total from Series A and B funding. It did reveal that it has shipped “millions” of boxes to date, which can help to explain why 2,800 of its 4,400 employees are “stylists,” who use data to help choose what goes in every box.
While most of those 2,800 stylists do not work full time, they are W-2 employees, meaning that Stitch Fix deducts payroll taxes from each pay check and offers benefits like 401K and health insurance to those who work a certain number of hours a week. (It’s a point of pride for the company in light of the legal issues faced by other startups — such as Uber and Postmates — that offer similarly flexible hours but treat workers as independent contractors, which means that they are not eligible for benefits.) As traditional brick-and-mortar retail jobs decline, Stitch Fix offers something arguably better: the opportunity to work remotely and one-on-one with clients. “It’s more than just an hourly job,” Lake says.
There is a reason the stylists are valued so highly. Data may be Stitch Fix’s calling card, but a human stylist is assigned to every single box, which explains the importance of rich customer feedback beyond scraping emails for certain words (natural language processing) or Pinterest boards for duplicate images. “The algorithm suggests an overall match, and then we as stylists comb through all of that rich information that she gives us,” explains Layla Katz, who worked as a personal shopper and stylist in the Bay Area before joining the company, first part time and now full time. “It can be really granular, like, ‘I just love a butterfly print.’ Or, ‘I’m swooning for off-the-shoulder.’”
Stitch Fix is confident enough in its approach to be expanding into other categories, including shoes, petites, and maternity. Plus-size has been a challenge — in particular, finding great brands that hit the price point Stitch Fix is after — but the push is underway. And the company has already begun testing men’s, which launches in the fall and will go head-to-head against the Nordstrom-owned Trunk Club, a personal styling service that expanded into women’s last year. (Colson says that while men are indeed less likely to use Pinterest boards to help a stylist along, early feedback suggests that they are willing to offer plenty of information about things like product preferences and body hang ups.)
Whether or not Stitch Fix will indeed replace trips to the mall for a significant number of Americans remains to be seen, but the ambition is there and the company certainly offers a lesson to other brands and retailers about the importance of understanding and focussing on your competitive advantage. “We’re not going to be better priced, or faster shipping, or a better brand,” Colson says. “So we have to be really good at relevancy.”