In 2015, Leah Ingram was about to embark on a media tour to promote her new book. She needed clothes for the tour but didn’t want to spend $100 an hour for a personal shopper. So she went with a less costly alternative: Stitch Fix, a style subscription service that sends custom-curated boxes containing five garments to people’s homes for a $20 styling fee.
Once she tried her first style box, Ingram was hooked. She says at this point, about 80 percent of the clothes in her wardrobe come from subscription services. Every three months, a new box — called a ‘Fix’ — arrives by mail, and she gets regular shipments from Nordstrom’s Trunk Club too, “way too friggin’ much.”
In addition to price and home delivery, there’s something else that differentiates subscription services from in-store personal shopping: Stitch Fix and other leading brands combine real-life stylists with Artificial Intelligence.
Why else would Stitch Fix need a Chief Algorithms Officer? His name is Eric Colson. In an interview on the podcast Data Camp, Colson said A.I. is woven into every aspect of the e-commerce company, from inventory and forecasting demand to selecting and designing clothes. (YR Media requested an interview with a Stitch Fix stylist and data engineer and the company declined.)
Stitch Fix clients start by completing a Style Profile. When I decided to try it I was prompted to answer questions about my location, profession and body type as well as fabric and price preferences. I clicked through Pinterest-style collages of outfits. When asked, “Is this your style?” I clicked “no” on a collage with a floral maxi dress and “yes” on a collage with a biker jacket.
Next, the client is matched to a nearby warehouse with relevant inventory. The company matches the client with a stylist using information from the Style Profile as well as data on a stylist’s personal taste and availability.
The algorithms weed out and select potential items to recommend to a stylist, who makes the final choices based on the client’s profile, their own judgments and what’s available in stock.
From a consumer standpoint, an actual stylist being part of the process is extremely significant, according to Jessica Ogilvie, a professor at Marquette University.
“It’s not robots you’re going out to impress when you’re going to work, it’s humans,” Ogilvie said.
Stylists offer more than last-ditch quality control. Ogilvie said they provide clients validation that their preferences were considered by an actual person.
“What consumers truly want today is not unlimited options but expertise,” Ogilvie said.
But can a combination of algorithms and human taste accurately predict style?
For Ingram, the writer about to go on tour, the answer is yes. Sort of.
“Nobody gets my style 100 percent, but the 30 percent or 40 percent that they’re getting is good enough that I will keep using the services,” Ingram said.
“There’s a really nice mix of human and machine in there,” Shanahan said.
It can take a while to get there. While style box companies tout themselves as time-savers, Shanahan said there’s a learning curve — for clients and the machine. After receiving a style box, clients are prompted to provide feedback, which is then used to inform future selections. It can take time for enough data to accrue for the style service to grasp a client’s individual likes and dislikes.
Shanahan said one time, he tried out a different subscription service and was too specific about what he didn’t like in the first delivery. As a result, he said he never received another box.
“You have to be willing to make that time and investment. If you give up after one, it’s not going to work.”
On a Reddit board about Stitch Fix, one commenter with the username tequila_mockingbirds asked if it is normal for Fixes to be initially disappointing. “The dress especially [was] just immediate revulsion in person and two sizes too big and ugh,” tequila_mockingbirds said.
Another commenter reminded tequila_mockingbirds that at the very least, Stitch Fix is still less effort than going to a mall. “My friend who works on their data science team says to give it about 6 fixes and to give as much feedback as you can,” the user said.
Still another commenter said that tequila_mockingbird’s experience is not an anomaly. “It’s totally possible for you and the service to be a bad match, TBH, so don’t doubt yourself,” the user said.
The day my Stitch Fix arrives, I run downstairs and find that the box is already on the kitchen counter.
“Your Stitch Fix came!” my housemate says.
I take the box up to my room and rip it open.
Inside I find: a pair of distressed bootcut jeans (Judy Blue, $58), a snakeskin printed kimono (Emory Park, $34), a black and white striped jumpsuit (Kaileigh, $58), a burnt orange blouse (Lush, $40) and a white V-Neck knit top (Market & Spruce, $28).
If I keep any of the items, the $20 styling fee is waived and I will only pay the price of the items I keep.
With this in mind, I try on everything, lingering on the cool and comfortable black and white jumpsuit. However, the kimono, jeans and blouse are respectively too feminine, too tight and too boxy for my taste.
And then I try on the white V-Neck knit top. As I stand in the mirror wearing it, I laugh. It is the plainest, whitest T-shirt I have ever seen in my life. In its simplicity, it is almost flattering. I ask myself how something so unspecific, so void of personality could be a reflection of my — or anyone’s — personal style. After filling out such a comprehensive survey eight days earlier, I am almost offended.
Should I keep it, just to have? I ask myself.
The next day I repackage the clothes in the return pouch they came with, take the pouch to the post office and walk up to an employee at the window.