As convenient as it may be, shopping for clothes online has one major pitfall: finding the right size. Retailers know it: E-return rates can reach as high as 40% as consumers take guesses on the correct size and toss items in their virtual shopping cart. That’s if they even bother to click “buy” amidst the uncertainty. It’s a sticking point for retailers that one startup hopes to fix.
True Fit, a retail software startup that just raised $10.6 million in equity funding, uses data analytics to take the guesswork out of online shopping. Well-known department stores and brands, like Macy’s, Nordstrom, and Guess, use True Fit’s algorithms on their e-commerce sites to help boost sales conversion rates and reduce customer returns. And in a quest to reduce retailers’ online return rates and boost sellers’ sales conversion numbers, the software company gets a lot of help from the field of data science.
“There’s no real easy way to figure out what size you should pick across brands and different styles, and especially when you don’t have the benefit of a dressing room,” says True Fit cofounder Jessica Murphy, who was a buyer for a large department store before starting True Fit.
To get a global perspective on how clothing sizes from different brands compare with one another, True Fit gets proprietary fit data from its more than 1,000 brand partners. So, if True Fit needed to understand how Levi’s intended a pair of jeans to fit a customer, it could just look through its database for the specific fit information that Levi’s provided.
“It’s not just a technology solution; It has to also be a relationship solution, as well as a product solution,” says Jeff Putz, True Fit’s vice president of engineering. The more brands and retailers the company partners with, the wider variety of recommendations True Fit can make to its clients’ online consumers.
When a user is ready to select a clothing size on a product page, he can choose to access or set up his True Fit profile. Setting up a profile requires him to know his height and weight, as well as the size and brand of his favorite piece of clothing. In less than a minute, True Fit’s algorithms spit out a size suggestion, with a rating that indicates how happy the user would be with the item. But what makes True Fit so easy to use is the last piece of data that the user provides about himself–his favorite item of clothing.
“The preference data is really the key to getting it right,” Murphy says. It is even more important than getting his exact measurements. With this one piece of data, True Fit’s algorithms can keep learning more about the user over time, all the while tracking the user’s sales and returns data. Murphy likens True Fit to the music-discovery app Pandora.
“It’s less about getting people into the right size, which is obviously very important, but it’s more important to get them into the styles that they’re going to keep,” Murphy says. The more users use their True Fit profiles with different brands and items, the better of a picture True Fit gets of what types of styles the user is likely to keep or return.
On average, True Fit has helped its clients reduce their online return rates by 10% to 50%, depending on the category. Lord & Taylor improved its sales conversion rate three times over after implementing True Fit.
To crunch all the numbers from both consumers and retailers, True Fit’s engineering and data science teams work together. Engineers participate in the data exploration with the data scientists, and the data scientists take part in the construction and delivery of the coded algorithms. Putz describes this collaboration the best way to innovate.
Both teams use the usual suspects of data science tools to sift through all of the data points that it gets. Among them are Hadoop, to process datasets in parallel, and R, to visualize and analyze them. The teams store all their data in the cloud, terabytes of it on secured, third-party servers.
“We’ve found that we can get the best results by using a variety of techniques in really clever ways,” says Zhidong Lu, lead scientist at True Fit. How it mixes and matches its algorithms is what makes True Fit’s approach to analyzing such large datasets unique. Some of the calculating techniques it uses are generalized linear regression, collaborative filtering, and quadratic programming, all of which are common tools among data scientists.
Exactly how True Fit’s number crunchers make their mix of calculations so “clever” is information that does not travel far outside of the company. But there is no question that how True Fit leverages its massive database and assortment of algorithms is improving retailers’ bottom lines.
“So much of what we do is what we’ve learned from using those tools and really making them fly in the real world,” says Putz.