When you think about the fashion and retail industries, the first images that come to mind are beautiful models on catwalks, suit tailors with measuring tapes and seemingly never-ending worldwide competition of who can build a better/bigger mall.
Technology, on the other hand, is nowadays associated with robots that can clean your house, cars that drive themselves and drones that deliver items your ordered within minutes.
Both industries have undergone a major transformation in the past 10 years but thanks to the e-commerce boom, there’s a lot more technology in the modern retail industry than you think. Every page you visit, every picture you look at, and every click is recorded and used in calculating the supposedly “relevant” products you should be seeing next. Once you visit those next products, the same cycle continues, where the records adapt and the “relevant” products are re-calculated depending on the smallest of gestures, like whether you stared long enough at the page or closed it right away. Every action you make contributes to a bucket of data collected, and bigger bucket means more patterns or trends that can be recognised by a computer. This is a discipline called machine learning and as futuristic as it might sound, it is growing to be so common in today’s businesses, that probably your neighbourhood baker uses it.
If you’re reading this, you probably know what Mr. Draper is and you’re asking yourself what could Mr. Draper be using machine learning for. The answer is: Everything. As a resident geek here @ Mr. Draper, let me give you a sneak peak into what goes on behind the scenes, between your first visit to our website to you getting a box of clothes.
Ever since the opening of Mr. Draper in 2014, we’ve been collecting data like there’s no tomorrow. We weren’t sure what to do with it at first, but we collected it anyways. We know your height and weight, your favourite colours and brands, what you bought and what your returned and why you returned them. We even know about that one time you were at a party and your pants tore.
All jokes aside though, from the moment your sign up @ Mr. Draper, we take you through 2 parallel journeys: one is the personal stylist that gets assigned to you (let’s call her, Dona) and the other is what gets done behind the scenes by our machine learning software (let’s call it “Don”).
Don has 3 tasks from the minute you sign up:
1. The raw selection: This is roughly what people do in a traditional shop. Use the information provided by the customer like height, weight, sizes, fits, likes & dislikes to pre-select and suggest items that you might like and that will fit you. Don goes one step further and organises and rates all the products in our warehouse, based on the likeliness of you liking the item enough to purchase it.
2. The body doubles: Next step is to find customers in our database that have overlaps with the customer in question, using sizes, as well as selections made and items purchased to arrange and rate them in order of similarity to the customer.
3. The finer selection: Use the body doubles line-up to iterate and adjust the raw rankings of items made in the first step, and account for people of similar taste (according to Don, at least). This gives us the first draft of suggested clothes, personalised for you.
Don might be able to process and produce a list of items in under a minute, but your stylist makes it her mission to get to know you personally, before making any decisions. At this point, Donna’s job is to get in touch with you, find out who you are, what you’re looking for, and whether you have a specific occasion you want to dress for.
Don and Dona work together and learn from each other all the time.
After a chat with the customer, the stylist goes through the initial list made by the algorithm, correcting mistakes, while also discovering clothes she hadn’t thought of yet. Don loves corrections and feedback and it doesn’t stop with the stylist’s changes. Once the package is sent and returned, the software adapts the list of clothes based on what was purchased, what was returned, as well as why it was returned.
This process is then repeated for every customer, and every package, whether it’s someone who just signed up or has been with us since day 1. It means that there’s not a single day where the machine learning software doesn’t actually learn. It’s constantly changing and adapting personal selections for each customer and, in doing that, every detail matters.
At the end of the day, that’s what makes Mr. Draper a truly “personal” service.