Consumers today are demanding a more personalized shopping experience. They expect that retailers and brands ― especially those they’ve already shopped with — will have insight into their preferences and tastes and that they will not market to them blindly. In apparel, perhaps more than any other category of product, this desire for personalization, of both marketing and product, comes with added layers related to sizing and fit — and sizing and fit preferences.
The ratcheting up of customer demands ― and retailers’ increasing ability to meet them — has been fueled in recent years by ongoing developments in artificial intelligence (AI) and other technologies. Using increasingly complex algorithms and increasingly vast gobs of available data available to them from a variety of sources (consumer past purchases, customer demographics, product attributes, global trends and much more), apparel companies are better able to market to the individual.
Consider Clarifai, an artificial intelligence company focused on visual recognition, which can provide a vast array of efficiencies to a business by sorting through content swiftly — say, filtering all content that includes cats or dogs, says founder and CEO Matthew Zeiler ― thus freeing up workforce time for activities beyond sorting through content manually. Alternatively, it can help increase basket size, for example, by providing users with visual search capabilities and real-time recommendations.
Speaking recently at the annual summit of the American Apparel & Footwear Association (AAFA), in a discussion entitled “A Million Markets of One,” Zeiler and two other panelists — Daragh Sibley of Stitch Fix and David Macy of Avametric ― provided insights into how some of these developments are taking shape — and it’s not always what you’d expect to hear.
Pushing your customer outside her comfort zone
One of the most surprising revelations came from Sibley, a data scientist at personal style service Stitch Fix. In case you’re not familiar with Stitch Fix, it follows a subscription model, which works like this: after the customer fills out a style profile, the company mails five hand-selected pieces of clothing to her door at regular intervals, as designated by the customer. She then buys what she likes and sends back what she doesn’t. Over time, using a combination of algorithms and real human stylists (and feedback from the customer), the company learns the customer’s preferences better and better, gaining greater accuracy and a higher conversion rate as it goes. That’s the general idea, anyway.
Sibley says that Stitch Fix’s data scientists “like to insert [them]selves into every business process where we can create value. Say, styling. How can we make sure we get the best five items to each client? We have another team focusing on ― how do we match up the right stylist with the right client? My team focuses on ‘more.’ How do we make sure we’re getting more of the items people love, or explore different styles?”
Stitch Fix has a unique business model. Because it is sending out items, you might consider it a push model, says Sibley. It’s unlike going to a brick-and-mortar store where clients make selections from off the rack. But that push model comes with a high opportunity cost. Every time the company sends something the customer doesn’t like, it is losing not only the sale, but the opportunity to have sent something she does like. That requires handling inventory a bit differently from the more standard practices of non-subscription, brick-and-mortar retailers. For example, if the company finds that a particular item is not gaining traction, it may choose to clear it out of inventory earlier rather than later, so that it can get new items in. After all, there are no opportunities for “markdowns,” and the company doesn’t want to risk alienating its customers.
So how many tries does it take before Stitch Fix gets it right? Here’s the rub. Says Daragh, “We don’t always want to get it right.” Wait. What?
“Early on, we want to get you some stuff you like. But maybe by box 30 we don’t want to just get you five items we know you’ll like. We may want to push you a little, get you to explore more. We want to create a sense of discovery,” he says. What Stitch Fix really wants is a customer to say, “’You know what, I would never have picked this out for myself, but I love it! I get compliments on it all the time.’” That’s a “sticky” experience that Stitch Fix can create for a client. It builds loyalty, and creates additional value. “We want to push outside their boundaries, knowing full well it may not create a sale,” he says.
Sticking inside a customer’s comfort zone
On the other hand, sometimes pushing a customer outside of his or her comfort zone is a path to frustration for everyone. David Macy, vice president of product, Avametric ― which allows brands to deliver accurate 3D renderings of their apparel (and accessories) on customizable digital body models (for web, mobile and AR) — for example, shared a story of setting up a body scanner in a retail apparel store. “You’d think that would be best,” he said, when it comes to capturing precise personalized body data that customers can use to get connected with garments that work best on their bodies. But what the company found, said Macy, is that people don’t want to be scanned. Being scanned requires undressing. It often requires wearing compression garments. Many people are not comfortable engaging in these activities in a store environment, he said. For the data the body scanner collects to be useful to the customer, it needs to be stored in the cloud, and “people are not necessarily comfortable with that, although they may be in the future,” he says.
Most interesting ― and most amusing — “it’s hard to compete with people’s images of themselves. They often look at their body scan and say, ‘no, that’s not me,’” Macy said, to much laughter from the audience — including from this writer. And then I remembered getting scanned in the white-light bodyscanner at [TC]2 about 15 years ago, and receiving the printout of my silhouette and measurements, and feeling a bit panicky, and saying, “that doesn’t look anything like me,” and quickly suppressing the memory of the entire episode for 15 years.
“So we’ve swung the pendulum the other way,” said Macy — gathering simple inputs, including height, weight, bust size, or simple archetypes: what shape you are and what size clothing you typically wear. “So we create an avatar that’s similar but not exact. That is much better for getting people go to the next step.”
As for the brands Avametric works with, the company finds that they are concerned with different things. Some are more focused on conversation rate, while others want to decrease the rate of returns. As is typical with many apparel online purchases, customers order three sizes and return two, which is of course very expensive for the retailer in terms of packaging, shipping and labor. “That’s no good,” says Macy. Being able to give customers the confidence to order one garment in a size they know will fit will eliminate a lot of wasted time, money and shipping and packaging costs from the e-commerce cycle.
Another goal of 3D visualization is reduce the time it takes to produce physical garments, “and perhaps get feedback before the garment is even created by allowing customers to visualize, digitally, garments on their own bodies in advance,” he says.
Jordan K. Speer is editor in chief of Apparel. She can be reached at [email protected]