Published on October 6th, 2014 | by Travis Korte0
5 Q’s for Romney Evans, Co-Founder of True Fit
The Center for Data Innovation spoke with Romney Evans, co-founder of retail software company True Fit. Evans discussed why it’s difficult to standardize different sizes for footwear and apparel across different brands and some of the insights True Fit’s client companies have uncovered using their data.
Travis Korte: Can you first introduce True Fit, what you make, and who uses it?
Romney Evans: True Fit is a retail software company. The company provides personalized recommendations powered by fit data from millions of consumers, a network of thousands of brands, and leading retailers. It’s currently being used by a lot of retailers including Nordstrom, Macy’s, Lord and Taylor, as well as a lot of direct brands like Guess, Arcteryx, and Sperry Topsider. We also have a pipeline of over 100 that will be launching over the next 18 months. And so in a nutshell, the company allows consumers who are shopping for clothing and footwear at a retailer to enter a little bit of data about themselves (no measurements), and then True Fit will give them a personalized fit rating and size recommendations for every style in that retailer’s catalog. We do all the data analysis on the back end and serve up a simple fit rating that indicates how well we expect that item to fit the user. And what we find is that it’s working to great effect.
TK: Have you seen any notable successes yet?
RE: We’ve had a lot of retailers sharing some of their results publicly. One brand has seen a sixfold increase in their conversion rate for users who have True Fit profiles, Lord and Taylor has seen a threefold increase, and Guess saw a 250 percent increase. The pattern you start to see is when users get more confident about what they see onscreen they’re more willing to make a purchase online. If you think of the market as a whole combining apparel and footwear it’s over a $1 trillion industry, and only ten percent is penetrated online. However, around 78 percent of in-store purchases are web-influenced, meaning the customer discovered the item online or they went online to get product information, or they went online to get inspiration, but they still made the purchase in a store. That’s still great for business, but what’s happening is that while the online space is naturally growing at a really high rate, for apparel and footwear it’s so far behind mature categories that there’s a consumer migration happening. Brick and mortar stores will still play an important role for consumers but the consumers are diversifying where they’re purchasing. More in-store purchases are starting to show up in other channels like web and mobile. This has been a reason why apparel and footwear sales have been suppressed online relative to other categories.
TK: How do you produce the recommendations?
RE: The software is built with a framework of machine learning algorithms and a robust set of data that allows us to deliver really high-confidence fit results. The way it works is we take three types of data. One is from our network of over 1,000 brands, including the product specifications like the measure, fabrication, and style attributes; basically, the blueprints of the products. From retailers we get transaction data, totally anonymized and not personally identifiable, but from that we can start to see patterns in return histories. And we connect those back to user profiles to see that users with certain attributes purchase items with certain attributes. That allows us to make recommendations to new users. Behind the scenes we are crunching tons of data, billions of transactions, millions of consumers, but for the user it’s very simple. You get a five-point fit rating about how well we expect an item to fit and flatter, as well as a size recommendation for that item. We can help people discover items more as they look at a catalog of items and help the best-fitting ones float to the top of the list. It’s a powerful enabler for consumers.
TK: Why is it that footwear and apparel brands haven’t managed to standardize size information on their own? What are the challenges associated with establishing baselines for fit?
RE: The reason brands have never been able to do it—and there have been people who have advocated for it—is that fit is a source of competitive advantage for these brands. Brands optimize their fit to who they believe are their core customers. This is why you can go to Abercrombie & Fitch and a size 6 there will be different from a size 6 at a Chico’s. They’re dealing with different customers. There’s also an element of vanity sizing, meaning strategically sizing down their sizes to appeal to the vanity of the consumer and make them feel like they are still a size smaller than what they might be historically. That definitely happens. Brands are sophisticated with regard to size strategy. They size in a way that maximizes revenue and profit. And people tend to be loyal to brands when they have positive fit experiences. This can be problematic for consumers as they shop across many retailers and many brands. You can think of True Fit as a translation service between all these different size systems. If we know what you’re wearing in one brand and we know some basic attributes about you the consumer, we can give you recommendations about other brands where you might not have experience that will be highly accurate. Because we’re using a machine learning approach, those algorithms are getting smarter over time, both for the individual and also for the population. It’s a constant learning process because the styles are turning over all the time, so it has to keep learning and re-learning.
TK: Have you found any other applications for this data? For example, have you talked to researchers about how they might be able to use the information?
RE: Yes. One brand we worked with found that there were groups of consumers using their site that were underserved. They made a recommendation to their product team to consider building two new fits in their anchor jean line to cater to this group of users that were motivated and interested to shop their site and were buying but were returning items disproportionately and not having successful purchases. That’s an area where they can use the insights from this information to improve their product suite. Another example is a prominent UK brand transitioning to the U.S. market that wanted us to do analysis on their U.S. system. If you index their size system, theirs were sized a lot lower than competing U.S. brands. They now have this information to make strategic decisions. They might decide they want to differentiate themselves and maintain the smaller size and just tell customer service people how to represent the product. Or they might want to resize those products to match up with U.S. benchmarks. We don’t share any brand information with other brands, since that’s held in confidence, but there is basic benchmarking we can do that aggregates information. There’s also a plus-size denim brand that approached us. We identified that their grading rules, which are the rules that define how the measurements of an item changes as you size the item up and down, needed to be tweaked to reduce returns at the lower end of the sizing scale. If they applied that to their whole business, we estimated about a 1.2 million dollar savings for them.