Stitch Fix: Mastering Personalization Through Data Science
Ever wished that you had a personal stylist on call to deliver EXACTLY what you want without having to shop online or in a store? You’re in luck. Four year old San Francisco based, Stitch Fix is the “personal styling service tailored to your taste, budget and lifestyle that helps you look and feel your best everyday.” This is quite a lofty customer value proposition for a company operating in a competitive and fickle marketplace but Stitch Fix managed to become a $150 million company after only 3 years in operation….here’s why:
Great Customer Experience
When a customer signs up they are prompted to fill out an extensive “style profile” which is about survey about everything from body type, height and weight to upcoming life events, number of dates attended in a week and spending limits. The customer is also asked to rate clothing styles and to select favorite colors and patterns. Consumers also have the option to share a link to their social media account and a Pinterest board illustrating style preferences. Once complete the customer pays $20 and receives a box 
containing 5 items, a photo detailing how to wear each style and a note from a Stitch fix personal stylist about a week later. Once the customer tries on all of the items she logs back onto the Stitch Fix website to check out. Should the customer decide to keep 1 or more of the items, the $20 payment is used as a credit against the purchase. Otherwise, the customer can use a prepaid mailing envelope to return unwanted items and leave feedback about what went wrong. If the customer decides to keep all 5 items she receives a 25% discount on the entire order.
The Element Of Surprise
In doing the research for this blog post, I got pretty carried away reading reviews and looking through selfies of women sharing their newest “fix”. Stitch Fix has managed to create this cult following among consumers mostly because of the surprise element associated with each delivery. Customers never know what they are going to get. It’s like opening a Christmas gift every week that only gets better. The idea that consumers only pay $20 for the attention of a stylist and some mysterious algorithm thingy only adds to all the hype.
And Of Course… A Data Driven Approach
The secret sauce behind Stitch Fix’s success is the company’s focus on data. Called the Pandora of fashion, Stitch Fix uses the data obtained from its consumers to create a completely personalized experience. The data collected is used to inform Stitch Fix stylists of the specific likes and dislikes of the customer in real time. The more a customer uses the service the more accurate their shipments will be. Stitch Fix’s ability to collect data on the items that are not selling sets the company apart and makes it an invaluable partner to buyers as well. In traditional retail, there is little data on why certain items sell while others don’t. In addition to being another retail channel for brands, Stitch Fix offers much needed data to its brand partners. This in combination with the company’s ability to deliver a convenient, personalized customer experience makes up Stitch Fix’s core competitive advantage and creates material barriers to entry.
For more details on how Stitch Fix’s algorithm works check out this link: http://multithreaded.stitchfix.com/blog/2015/03/11/word-is-worth-a-thousand-vectors/
Great post! I have a friend who uses Stitch Fix, and I’ve heard really interesting things about the service. I wonder though how they obtain the supply of clothes, and if that has an effect on the types that they ship to customers. Are they partnering with stores or are they sourcing their clothing directly from the large suppliers?
Thanks for your post! I am also a user for these fashion in a box. However, one pain point I have is the fact that sometimes they don’t get my style / size right for the first time. Do you know if Stitch Fix design any mechanism to correct the data that we key in initially (well, sometimes we tend to under-report our waist size) to provide a more accurate experiences for customers?
Great post! I’ve never tried Stitch Fix personally but my sister has, and I remember her being a little put off by her first few experiences, where she felt like a lot of what she was being sent was in fact the opposite of what she had mentioned in her initial style preference survey. I wonder how they’ve managed to improve this over time (e.g. maybe larger inventory selection or more sizes, etc.) and if they’re actually using the “bad” boxes to measure how likely a customer is to actually return something even if it’s not necessarily their size. It could be a cool way to test customer stickiness and the ability to push through inventory to unlikely customers who may not have purchased those items individually at a store.
Thanks for your interesting post. One thing I’ve always wondered about with these curated services is how they strike the balance between sending you things that are “exactly you” that you always will like vs. “stretch pieces” that you wouldn’t necessarily pick out for yourself and might work every other time. I could imagine that if you were a loyal, reoccurring customer that eventually the data-driven styling will converge to boxes that look exactly like clothes that are already in your wardrobe (either ones that StichFix sends you or ones that you bought yourself). One of my good friends tried the service and ended up sending back all of her clothes because they looked like clothes she already had. I wonder if these algorithms can get too good at predicting your style and don’t take enough risks.
What I think is so interesting about the Stitch Fix model is how they blend the data and algorithms with a personal touch. No customer wants to feel like a mathematical equation has picked out clothing for them, and an algorithm in and of itself doesn’t create a very sticky customer relationship. What I think is so valuable about Stitch Fix is that the company initially uses an algorithm (that’s based on a combination of stated preferences in the style profile with derived importance of what a client has kept from previous boxes — and learns over time), but then relies heavily on a remote group of stylists to actually curate a box. Once a client requests a “Fix”, the stylist receives the profile with a large list of algorithm-selected merchandise from which the stylist selects 5 items. The system acknowledges that algorithms may not be able to deliver a full CVP, and a human touch is sometimes still necessary.