IBM's Watson is ready to help pick your wardrobe

The North Face is using advanced AI to help customers find the right gear

The IBM supercomputer Watson can win Jeopardy games and power intelligent toy dinosaurs, but now the advanced software is helping customers of The North Face pick out jackets and outdoor gear, just by having a conversation.

On Monday, the sporting goods retailer unveiled a new way to search its online catalog that utilizes the "natural language processing" seen in digital assistants like Apple’s Siri, Microsoft’s Cortana or Amazon’s Echo. Rather than filtering a list of products to narrow a selection, customers type questions and criteria into the interactive site, much like they would query a sales associate at a brick-and-mortar store.

Fluid, an ecommerce software company, designed the platform utilized by The North Face, called Fluid XPS (expert personal shopper). Powered by Watson’s algorithms, it parses the requests and determines what other information it needs before recommending an appropriate product, then prompts users for that input.

"This partnership provided a way for us to take in context what consumers were looking for in their next adventure — where they were going, when they were going, what type of activities they were doing — and being able to serve up the right product that we offer for that adventure," said Cal Bouchard, director of ecommerce for The North Face. "We thought it would be a really different way to shop online. If you think about the history of what e-commerce has been and is today, it still really is a white background with a grid of products."

Fluid XPS is not. The white text set against the black background asks for a place and a date. Mountain climbing in January in Colorado prompts a few questions.  Let’s say, my aunt is going on the trip, and she wants an orange jacket (easier to see, after all).

Fluid XPS uses Watson to calculate the expected conditions: 20 degrees Fahrenheit and 11 MPH winds, then recommends six different women’s jackets that are available in orange, or at least bright red. Review the details of the conversation to see the criteria. "Based on what you’ve told me about your trip to Colorado in January I have selected jackets that are designed for freezing, light wind, aunt, mountain climbing, snow, orange."

That’s the kind of nuance not available from a traditional e-commerce site. A typical filter wouldn’t show jackets that don’t come in orange. But maybe the customer isn’t wedded to that color. It also understands that an aunt (or a daughter or a sister) is a woman. XPS provides alternatives and tries to present the best match.

The software is still in beta, so there are a few snags. After announcing the temperature would be 20 degrees, the site then offered a choice between snow or rain, and heavy, midweight or light jackets — selections that would have been obvious to a human.

Another benefit of the system may not be apparent at first. Products with varied technical specifications can be difficult for consumers to choose between if they lack the expertise. That can be mitigated by in-store personnel or by live chat online, but automated systems don’t offer much help. "Consumers need to be guided. They don’t even know which questions to ask," said Neil Patil, president of software business at Fluid. "They need a system, especially online, where it’s asking you questions that you can just express in your own words and not have to understand different technical terms."

In the future, he expects the XPS platform will be used for products like electronics where technical knowledge can be a barrier to a satisfying shopping experience. Now that the initial infrastructure is in place, brands looking to take advantage of the platform could be up and running in as little as 90 days, Patil said. An unnamed consumer products brand in the healthcare and beauty industry is currently in beta testing, and early in 2016, Fluid will begin working with new verticals, starting with electronics.

But rolling out a virtual shopper takes some preparation. "There’s one things brands may not have ready themselves, and that’s understanding the granular purchase criteria of their products," Patil said. A brand’s catalog needs to be fed into the XPS system item by item. That involves breaking down products into searchable criteria that consumers may want, and categorizing them in all the different ways consumers might ask for it. That’s much more complicated than a typical information tag or product information list.

Customers on The North Face site can ask for jackets that are packable, good for skiing or hiking or both, or just used for a commute. With or without pockets. Down or synthetic. Sometimes it’s just a keyword or a tag, but often it’s more complicated, and in future products — say, high-definition televisions — the old ways of classifying features won’t work. If a consumer doesn’t understand what HDMI is, they won’t include it in their question, but what tags apply to something like "good for having friends over to watch the game"? The input is becoming more "human," but filing systems are lagging far behind.

"[Customers] are used to using a filter or an onsite search or using keywords, but it’s hard to get out of them how they want to interact," Bouchard at The North Face said. "Asking them about how to use something that isn’t around is really hard."

Subscribe today for just $116 a year

Get the very latest news and insight from Campaign with unrestricted access to , plus get exclusive discounts to Campaign events

Become a subscriber


Don’t miss your daily fix of breaking news, latest work, advice and commentary.

register free