fabernovel loader

Jun 19, 2017 | 3 min read


How AI Enabled Stitch Fix to Create a Mass Customization Service

Tom Morisse

Research Manager

Too often pitted against each other as substituents, humans and AI algorithms can prove highly complementary. The proof by Stitch Fix, a personalized fashion shopping service.

The pitch – how AI is leveraged


Launched in 2011, Stitch Fix offers a personalized shopping experience for fashion. Based on a 10-minute profile filled out by each client, a stylist selects 5 clothing and accessories items adapted to the specific tastes and budget of every user. Selections are then sent, regularly or on-demand, to the clients to try products on at home. Back and forth shipping is free, and customers are only charged for the items they decide to keep. Finally, they are free to send feedback to their stylist.


While such a shopping model is not new (for instance, Trunk Club was founded in 2009 with this personalized home experience idea), Stitch Fix benefits from a key advantage in the way it leverages AI to help stylists in the selection process. Its machine learning algorithms surface relevant items for the customers at hand, then stylists make the final decision.


What is truly interesting about Stitch Fix is that the company started building its data science and machine learning capabilities right from the start – its Chief Algorithms Officer Eric Colson was hired back in 2012. Out of its more than 5,000 employees (including 3,000 stylists and 1,200 in warehouses), the Algorithms team is 70-member strong. This team is subdivided into 4 groups:

  • Client Algorithms: “Understand our clients, predict their demands, and optimize operations.
  • Styling Algorithms: “Meet the recommendation engineers: Where the art and fashion is quantified and shipped.
  • Merch Algorithms: “Mastering the science of clothing, inventory, and complex logistics.
  • Data Platform: “Build robust and powerful technologies to store, access, and compute data.



The impact – what AI brings


The right human / AI balance: Stitch Fix’s service merges the best of both worlds. On the one hand, algorithms are best at sifting through structured data, i.e. generating recommendations thanks to 3 sources of data:

  1. Customer preferences survey (e.g. size, style, budget)
  2. Product information (e.g. color, material, price)
  3. History of customer-product interactions (i.e. which products customers keep and which ones they send back)

On the other hand, stylists can easily use the unstructured data provided by customers, such as feedback messages or links to their Pinterest boards. In addition, they get enough time to focus on having meaningful interactions with the users, as the numerous letters shared on social networks by delighted users demonstrate:

Improving customer and employee experiences at the same time: the heavy reliance on AI enables Stitch Fix to do away with complex websites or apps. The low-touch customer experience thus frees up resources to improve the stylist interface. This results in greater productivity and more time to take care of clients.



Ripple effects – what could be next?


Ultra-low-touch experience: instead of sending selections on a ‘pull’ basis (customer-driven schedule or on-demand), Stitch Fix could take advantage of all the interaction data it has gathered so far to ‘push’ shipments to its users, predicting the right moment when they could need new clothes. The customer experience would thus be reduced to its essence: trying, choosing and interacting with a dedicated stylist.


Ongoing customer relationships that allow for diversification: algorithms will continue to improve stylists’ productivity, including to better keep track of customer messages. But one thing that Stitch Fix should not automate is the stylist-customer interaction part. Leaving its stylists plenty of room to empathize with customers and personalize their experience is the luxury that its data science prowess has made possible. On the contrary, a higher AI-driven stylist productivity should lead to even more conversations with the users. All the pieces of information gleaned that way could open the door to a diversification of products sold, extending from fashion retail to a lifestyle assistant. What if Stitch Fix sent home decor or food items inside its boxes? Or suggested events to attend? Given its stylist workforce is distributed across the whole US territory, it can develop a deep understanding of each customer’s very own context.

Interested in receiving every week a new episode on Artificial Intelligence Season 2 by FABERNOVEL?

This article belongs to a story
logo business unit


Distribute the future. Connect leaders. Change the game. We ignite ventures and transform organizations for the new economy.

next read