With Stitch Fix, users don't go shopping for their clothes. Professional stylists do the job for them and the personal shopping service ships the new clothes to their door.
The stylists aren't working on their own, though; they're using artificial intelligence (A.I.) and a team of about 60 data scientists.
That combo is behind the success at Stitch Fix, a San Francisco-based online subscription and shopping service founded in 2011.
"I think it's the single most salient aspect of our company," said Eric Colson, chief algorithms officer at Stitch Fix. "Our business is getting relevant things into the hands of our customers. This is the one thing we're going to be best in the world at. We couldn't do this with machines alone. We couldn't do this with humans alone. We're just trying to get them to combine their powers."
Stitch Fix, a company with about 4,000 employees -- 2,500 of them working as stylists -- has amassed a following among busy women -- and as of February, among men. That's when the company launched a beta service for guys, with a full public launch scheduled for this fall.
Stitch Fix is aimed at people who either don't enjoy shopping or simply don't have the time to go to a brick-and-mortar store or cull through endless online pages of shirts, pants, sweaters and jackets.
Users start by filling out an online style profile. Do you like blousy or close-fitting tops? Favorite colors? Are you more urban hipster, Sex in the City chick or Bohemian? Do you prefer jeans over dresses?
Stitch Fix stylists, both human and machine, handpick a selection of five clothing items and accessories that fit each client's taste, budget and lifestyle. Clients keep what they want and return the rest.
Colson, who was a vice president of data science and engineering at Netflix before joining Stitch Fix in 2012, noted that the company was using an algorithm for basic criteria filtering. If a client was a medium, it would filter out shirts that were a large or small. If she didn't like the color yellow, those items would be excluded.
With his experience at Netflix, which bases much of its business on recommending movies and shows to users, Colson knew Stitch Fix could do more by using machine learning.
"A product manager at Netflix used to say if we were really bold, we wouldn't present five or so recommendations. We'd present one and if we were going to do that, we should just play a recommendation when the user came online," he said. " Here is Stitch Fix being so bold as to say, 'Don't worry about picking out stuff. We'll do that for you.' That was exciting and bold. Could that be done? Is that possible?"
Later in 2012, the company got its first machine learning algorithm, which was designed to get smarter the more data it handled.
"We've been able to augment human judgment with machine algorithms," said Colson. "We have to combine machines and expert humans. It turns out it works better than even I could have thought."
Today, the company has hundreds of algorithms, like a styling algorithm that matches products to clients; an algorithm that matches stylists with clients; an algorithm that calculates how happy a customer is with the service; and one that figures out how much and what kind of inventory the company should buy.
Stitch Fix also has an algorithm that learns from images so it can check a client's Pinterest pins and learn what styles she's favoring even if the user has a hard time articulating it in an online form or in comments.
The company, according to Jeff Kagan, an independent industry analyst, is likely ahead of a trend where machine learning moves into the enterprise.
"I think this is just the beginning," he said. "Many new business models will form with A.I. as the center of their universe.... We have to remember this is advanced and cool, but it is still very early in the process. This is the Model T for the A.I. revolution. It will get bigger and better, year after year."
For today, Stitch Fix's human stylists, most of whom work remotely, use an interface where they can see all the information the company has about a client - measurements, preferences, desire to take risks. The stylists use the interface to see comments about previous outfits they've tried, and notes on whether they have a big date or a wedding to attend and need something new to wear.
The stylists also use the interface to see all the images of products the algorithms have recommended for the client.
That, said Layla Katz, a lead stylist with Stitch Fix, is incredibly helpful.
"I quickly realized the tool was my new BFF," said Katz. "It gives me confidence when my creative eye is saying this is a match and the science is saying the same thing.... How they come together is the magic."
Katz, who has been a stylist with Stitch Fix for two and a half years, said she likes working with the artificial intelligence. It makes her job easier and gives her more time to be creative.
"When a client fills out a profile and is ready to be styled, we are able to see what the algorithm is suggesting based on the data collected from her profile -- everything from sizing to location, geography, body type, fabric preferences, colors and pattern preferences," she explained. "It helps to not have to worry about the broad strokes of what a client does not want. Then we can make creative decisions about what will fit her body and her lifestyle."
Without the A.I., it could take stylists weeks of working with a client to know what works.
But don't get Katz wrong. The human stylists make the creative decisions.
"Once we build a relationship with a client, we get smarter every time [we style her] and the algorithm gets smarter every time," said Katz. "When a client decides which items to keep or send back, she can go through her profile and let us know item by item if she liked the fit, the price, the quality. That goes into the algorithm and helps it suggest more for the next time."
The algorithms, for instance, take in customer comments and feedback and can easily and quickly calculate how customers feel about certain products and styles. That flood of information would be overwhelming for a human.
"Humans are better at interpreting the tone and meaning of textual feedback, but they can't do many," said Colson. "After a few thousand, they might get really bored or have a hard time distilling it down. Machines can hold in their memories far more than we can.... The challenging part would be not using algorithms. It would be scary out there. It would be a leap of faith to put your finger in the air and make the best guess about what people will buy. That would give me anxiety. We have the benefit of data."
That ability to digest and analyze so much data could be what makes Katz such a fan of working with an AI system. Elsewhere, many people are afraid of smart machines and robots stealing their jobs, said Dan Olds, an analyst with The Gabriel Consulting Group.
"It's a little surprising that the stylists don't see A.I. as a threat, since most people seem to think that A.I. will eventually take over their jobs," he said. "But I believe we'll see A.I. assisting people in almost every job, particularly those in which there's a lot of data that has to be processed. When you think about the clothing industry, a good A.I. can help winnow down the universe of clothing and accessories to help the stylist do their job better and quicker."
"I am a huge lover of all things algorithm," she said. "I'm a real fan girl. It's an insane genius what the algorithms bring."
Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.