News Stay informed about the latest enterprise technology news and product updates.

Fashion tech startups use data science to build virtual dressing rooms

Virtusize, Clothes Horse and LoveThatFit are trying to solve a burning question for online shoppers: But will it fit?

Shopping the latest fashions online is convenient for the customer, but it has created a new headache for retailers -- and a new niche market for entrepreneurs.

You can never try on clothes when you're shopping online, so you can never determine the fit of the garment you're looking to buy," Rasmus Thofte, head of North American sales at software-as-as-a-service provider, Virtusize, said during a panel discussion at the recent Strata + Hadoop World. That may not seem like a hard data problem, but size in the fashion world varies from brand to brand and, therefore, is not necessarily indicative of how the garment will fit.

"That's intentional because if someone knows Levi is always going to fit, a customer will go back again and again to that brand," said Liza Kindred, founder and CEO of the fashion tech consultancy Third Wave Fashion in New York City and the panel's moderator. "There isn't a lot of agreement across brands about what a specific size is, which, for consumers, is a frustration."

The lack of any kind of virtual dressing room contributes to a high rate of return, which ranges between 50% and 80% depending on the source, Kindred said. It also contributes to high shopping cart abandonment rates, which range between 60% and 75% depending on the recent study, according to statistics compiled by Baymard Institute, a research institution based out of Denmark that wants to improve the online consumer experience.

Finding the right 'fit'

Virtusize, based in Stockhom, Sweden, is one of a handful of fashion tech startups trying to solve the "fit" problem -- and it's doing so in a personalized way. Inspired by old-school eBay resellers and the tailors who replicate favorite pieces of clothing for clients, Virtusize asks the user to provide the measurements of a favorite dress shirt or pair of jeans. The measurements are used to create reference garments so that, when a customer is shopping online and wants to take a closer look at fit, he can layer a silhouette of the garment he's interested in buying over the silhouette of the garment he already owns.

Ramsus ThofteRasmus Thofte

"You can call the sizes whatever [you want] because all you're getting is a visual representation of those two garments," Thofte said.

But even if sizes were standard, the fit problem would persist. Take the data variety challenge that exists for dresses, which Thofte pointed to as the most difficult item to fit. "There are so many different types of dresses, which require a lot of silhouettes. There can be hundreds of different types of dresses within one single retailer," he said. And, even if the shopper knew exactly what size to order, she still might not know about drape, or how the cut and fabric of the dress might work for her.

David Whittemore, co-founder of the New York City-based startup Clothes Horse, agreed. "The silhouette problem is totally relevant," he said. As is the difficulty with determining the fit for stretchy fabric, he said.

David WhittemoreDavid Whittemore

Clothes Horse, which was acquired in December by, has built a data-driven recommendation engine that leverages data from both consumers (body type, favorite brands, fit preference) and clothing manufacturers to create a "Switzerland, neutral third-party database of actual fit data -- chest, waist, hips, sleeves -- for every garment, across every brand in our database," Whittemore said. Proprietary-built algorithms that take things like fabric into account when making a recommendation underpin the engine. At the time of the panel discussion, Clothes Horse was working in 100 different product categories -- from down jackets to yoga pants.

"Those are two very different garments to get the right fit of," Whittemore said. "And the algorithms we designed and the data we were looking at had to map to those products."

The customer is always right

Garments aside, for the service providers to be successful, they also have to know their customers -- and their customers' sense of style. While retailers and designers may label the garment a particular category or style or fit, the real expert on categorization is the shopper. "If you're the L.L. Bean/Land's End shopper, you're different than the European designer brand devotee. Knowing that about the shopper lets you create more accurate recommendations," Whittemore said.

Gina MancusoGina Mancuso

The companies track individual customer data, although Virtusize users can access the service without creating an account. For LoveThatFit, a startup based out of Charlottesville, Va., that kind of data is vital to get closer to the customer. "We're collecting all of that data to understand their preferences," said Gina Mancuso, the startup's founder. "Because preferences are big. I might like a baggier fit, you might like a tighter fit. And now we can start to give recommendations for other garments because we understand their personal style."

LoveThatFit, which is currently in a private beta that it hopes to make public in March, comes at the fit problem a little differently than the other two. The company asks consumers to upload a full body image taken "in snug or tightly fitted clothing," to enable customers to try clothes on virtually. The startup uses proprietary-built algorithms "to adjust for tilt and distortion in the picture, while accurately finding the points of fit on your body," according to the startup's website. And, the platform is a social one, enabling users to share potential purchases within a shopper's network.

Next Steps

Fashion and data science intersect in this Q&A with Gilt's data scientist, Igor Elbert, and his experiments with pre-emptive shipping.

Dig Deeper on Enterprise business intelligence software and big data