Stitch Fix: Your Personal Shopper

Data analytics is at the backbone of Stitch Fix’s business model – a personalised online shopper for men and women. The company has experienced tremendous growth over the last few years.

Through the use of data analytics, Katrina Lake, CEO of Stitch Fix, is transforming the way customers shop. Stitch Fix is a personalized shopping website, where data scientists pick out garments for customers based on a complex algorithm and have it delivered directly to a customer’s doorsteps.


How does it work?

After a customer sets up an account on the Stitch Fix website they are led through a comprehensive survey about their style tastes (refer to figure 1). The data from the survey is then fed through a recommendation engine which takes a first stab at picking out relevant items for the customer. The items are then forwarded to a human stylist who picks out exactly five items to be shipped in a box straight to the customer. Once the customer receives their unique box within three days they are charged a $20 styling fee and have to decide whether they want to keep the items in the box or return it back to Stitch Fix. Once this is done, customers are asked for feedback so Stitch Fix can continue to build on the dataset they have for each customer.


Figure One: Sample of survey question


How has Stitch Fix performed?

Stitch Fix has grown to over three million active users since its inception in 2011. There 2019 revenues have grown to about $1.5 billion, translating to $488 net revenue per active client – a 9% increase from 2018. The company also boasts a 63% match score, the estimated probability that a specific client will buy a specific item.


How have they been so successful?

  • Integration of data science and human connection

The use of data helps in creating efficiency and effectiveness in picking out styles for clients, while the human stylists ensure that these styles are the right decisions. Having humans make the final decisions allows the company to add another layer of curation that the data may not have picked up on or add a little bit of surprise or spontaneity. This is especially true for such a personal experience like retail and especially if a client has a special request – i.e. what to wear for a specific event. Having the human touch augment machine learning helps to build trust in customer relationships.


  • Data to build on customer relationships

With each added interaction with customers, Stitch Fix gets better and better at predicting what their clients want. Their ability to continuously learn from customer feedback after delivering each box helps they develop a sustainable relationship with loyal customers. This also means that there will likely be low switching costs, given that Stitch Fix already has access to a large amount of customer data. This is evident in Stitch Fix’s above average 6-month retention rate of around 30% vs. other similar fashion companies.


  • Data science a critical component of the organization

The company has its own data science team of 115 individuals with its own Chief Algorithms Officer that reports directly to the CEO. This is important in signaling to the organization the important of data science in executing on their strategy. It is also important in making sure that all departments from operations to marketing interact with data science and utilize data to be more effective in how they conduct their business – i.e. to stock inventory more efficiently. It has also helped to foster innovation in the company, for example with the launch of hybrid designs, an in-house clothing brand conceived to fulfill a product gap in the market that was picked up using data analytics.



What are some challenges?

  • Stitch Fix owns its inventory

Stitch Fix has a business model where it owns all of the inventory it pushes to its clients, meaning that any incorrect decisions can prove costly to the company. It is important that the company optimizes its inventory management to make sure they have enough of a particular clothing but also to make sure their clothes are turning over at a fast-enough rate. This risk is mitigated by continuously improving on their algorithms and making sure they can attract a large number of users to build up their data collection.


  • Competition

Stitch Fix’s competitive advantage comes from developing a robust recommendation algorithm that can effectively match users with their clothing preferences. But what is to stop Amazon or another large retailer from investing in the required technology to build their own system? Especially if these larger retailers will already have a customer base to leverage. To combat these players, Stitch Fix will need to strengthen their relationships with users perhaps through increasing their supplier inventory or introducing some sort of loyalty program.


Word: 752







Hello Barbie – AI Making Children’s Dreams Come True


Nike – just Do it with Data science and Demand sensing

Student comments on Stitch Fix: Your Personal Shopper

  1. Really interesting to learn how they have blended data science algorithms and human talent to curate the best possible clothing selections for their customers. I agree that the threat of a larger incumbent entering the space is real but one clear advantage stitch fix has is each user’s specific clothing preferences, which will continue to grow. Traditional retailers or even Amazon is unlikely to have this level of data. One way to leverage this advantage is to build their own clothing lines based on their preference data or to partner with traditional retailers to utilize the vast amounts of data they have.

  2. Interesting post, thank you!

    I was intrigued to learn that Stitch Fix still have a human connection as the last step in their process. Do you think Stitch Fix will ever have a “machine-only” option, where you have no human connection but the price is cheaper? They certainly have the wealth of data to make this a reality – especially for long-time customers.

    I would rather pay <$5 for the recommendation engine to pick the right clothes vs. $20 for a human. But maybe that's just me – I appreciate that some people would always require a 'human-touch' and may even get attached to their stylist. To make a machine-only option a reality, Stitch Fix would need more parameters to check the outcomes, which may make the system more complicated (e.g., checking to make sure you are not sending people shorts during the winter in Boston..).

    1. Totally feel the same about preferring to pay less and forego the human touch. I have a hard time believing that the presence of a human in every transaction is necessary from a QA or algorithm improvement standpoint – so I wonder if it just exists to give customers peace of mind at this point, and Stitchfix believes it is necessary to convert customers?

  3. Great post! I didn’t know that there was a human component to their recommendation process, and i might be what helps them have a better retention rate than their competitors’. However, the fact that they own their own inventory might be a source of concern. I wonder if they’ll fall into the trap other companies that also own inventory have fallen into and start using their recommendations as a way to get rid of their excess inventory artificially pushing particular items. This may impact the quality of their recommendation and their customer satisfaction.

  4. Really interesting post on Stitch Fix! I am curious about the inventory management aspect. Specifically, there is obviously a huge financial risk by taking on this inventory. However, I was wondering if there was a way to do the model without this inventory. For example, if they transition into offline, would they be able to offer the Bonobos model and offer showrooms instead? And is there a way to translate this to online (perhaps only sending “samples”)?

    1. I totally agree with this and was wondering myself how big and broad the inventory is, not only in pieces, but in sizes. Looking at the sample of the survey question I thought that maybe it was not that detailed in the tailored approach, but more of a number of styles that fit most. I see that a good curation of style might come with partnering with many apparel companies and this would also help with the inventory cost and management.

  5. Thanks for writing about this! I remember once reading that one of their early hires was from the Netflix data science team and this created a culture building out robust capabilities in this area. For me this highlights the importance of picking the right early executives as they will have a lasting impact on the the DNA of your company. Based on your blog post, it sounds like Katrina made an excellent choice in this respect as data analytics has been key to Stitch Fix’s success.

  6. Thanks for sharing! I wonder how design might become a “victim” of a business model such as this one. As much data as it gets, the technology behind it would clearly become very successful at predicting similar items that the person might be suited for. However when you think of a real-life personal shopper or designer, their key edge is how they might innovate in design and propose something that the buyer might not have otherwise thought of.

  7. Great post! The Stitch fix model is truly innovative but I am bearish on the long-term sustainability of the business in its current form, having tried to use the subscription twice and returning all items on both occasions. I don’t feel the Human assisted Advanced analytics employed today truly captures the customer’s needs. feel the company will continue to have high-churn rate and higher customer acquisition costs. Although I do feel eventually the algorithms would become superior with more data and demographically diverse customer base.

Leave a comment