Stylefeeder Dissects Shopping Trends By Zip Code; Victoria’s Secret is Hot in Seattle, Hurley Beachwear is Tumbling in San Diego, Marc Jacobs is de Rigueur in Boston
Cambridge, MA-based Stylefeeder is bending more brainpower toward the study of online shopping behavior than virtually any other startup I know. The personalized product recommendation company has seven full-time employees, and five of them actually write code.
Recently, they’ve been coding up some cool algorithms that show site visitors a customized set of products depending on their location. Not only are visitors twice as likely to click on Stylefeeder’s product suggestions when they’re filtered geographically, says founder and CEO Phil Jacob, but by tracking who clicks on what, the company is gradually assembling a detailed, real-time picture of shopping trends in various regions of the country. Jacob shared some of Stylefeeder’s data with Xconomy this week—and it’s extremely interesting stuff.
By tracking the Internet addresses where Web requests originate, Stylefeeder is able to pinpoint which brands of shoes, apparel, and other items are gaining and losing in popularity in each region, down to the level of individual Zip codes. To cite one data set Stylefeeder sent us, site visitors from Seattle have been shying away from traditional, outdoorsy brands like Levi’s, Timberland, and Converse this week and toward more expensive designer brands like DKNY, Victoria’s Secret, and Kate Spade. Since Thanksgiving Day, DKNY has moved up almost 60 slots in Stylefeeder’s rankings for Seattle.
In Boston, meanwhile, The North Face, Nanette Lepore, and Diane von Furstenberg are down, while Lucky Brand Jeans, Moncler, and especially Marc Jacobs are up. San Diego shoppers seem to be losing interest in Hurley, Volcom, and Kenneth Cole, but are excited about Puma, Oakley, and Charlotte Russe.
Here’s the data in chart form:
While most clothing and accessory purchases are highly personal—Stylefeeder abounds with boots, blouses, belts, and bags—the company is also tracking items that site visitors are probably perusing as potential gifts for others. Here’s a rundown of the hottest gift items this week in Xconomy’s three cities:
Stylefeeder doesn’t compile all this regional data out of idle curiosity—ultimately, it helps the company show site users items they’ll like better.
While the company has spent years perfecting the collaborative filtering engine that underlies its personalized search service, that engine only works for regulars who have accounts and login names on Stylefeeder, according to Alex Nauda, Stylefeeder’s director of recommendation technologies. But many users arrive at the site from search engines or other locations, or never bother to log in. So Nauda and the other Stylefeeder programmers have put quite a bit of work into tailoring which products the site shows to these so-called “cold start” users—those for whom Stylefeeder has no preference information.
“We track what we do know about them, and the first, most dense, and richest thing is where they are in the world,” says Nauda. “I geolocate their IP address, which usually gets me down to the Zip code—not just which city, but which part of the city. Look at Woburn, Massachusetts, for example. Our user base there seems to be younger girls, like teens, who are shopping from home and go for more youthful brands like BCBGirls. Whereas our users in downtown Boston are older and more professional and would like maybe a Kate Spade shoe that’s much more expensive and suitable for an office. We track those metrics—what they click on and whether they return—and we use that to tailor results for other people who are hitting us from a cold start.”
It’s a powerful technique. Says Jacob: “In densely populated areas where we have a bunch of preference information, like New York and L.A. and San Francisco and Chicago, we’re able to double the click-through rates” for cold-start visitors. (Meaning that twice as many people click on Stylefeeder’s product images and are taken to e-retail sites where they can purchase them. Stylefeeder’s business model is to earn commission revenue on any purchases that result, as I detailed in a pair of posts back in April.)
As an accidental result of collecting all this regional preference information, Stylefeeder is gaining an increasingly deep, fine-grained, and immediate viewof shopping trends. In the past, the company hasn’t exposed much of this information to public view—in fact, the charts Jacob and Nauda shared with Xconomy are among the first bits of Stylefeeder intel to see the light of day. But Jacob says the startup is starting to think about how to make the information more available to e-commerce players who might find it valuable.
“We’ve known all along that retailers would be really interested in getting ahead of some of these trends,” says Jacob. “There are some people pushing us to package it up and sell it as a tool for people to play with, Alexa-style,” referring to Amazon’s Web traffic analysis subsidiary. “Just because of how we’ve been focusing on monetization, we haven’t done much yet. But the guts of the system exist already.”
So stand by to see whether Stylefeeder branches out into database marketing services. But meanwhile, what does all the city-specific shopping data really mean? Do Stylefeeder’s numbers reveal anything about the deeper sociological or economic trends affecting shoppers?
Nauda thinks they may. He emphasizes that his opinions are speculative, not scientific. “But just starting with Seattle, I think our data is indicative of something to do with the fall of the solid ‘Americana’ brands,” Nauda says. “Timberland, Converse, Levi’s–these are your standard American brands, whereas the stuff that’s rising is more designer-oriented–DKNY and Kate Spade. It’s telling, because the outdoorsy brands like Timberland are what you would normally associate with Seattle, but it’s the other brands that are on the rise.”
This week’s boost in click-through rankings for moderate-to-expensive brands like Victoria’s Secret, DKNY, and Kate Spade could also be signs of “a slight opening up of the economy,” Nauda says.
Stylefeeder will probably never be able to explain why waffle makers are hot in Boston, or why shoppers are gunning for Kalashnikov air soft BB guns in San Diego. But for e-retailers and their marketing consultants, the why may not be as important as the what. “There are, in fact, real patterns underlying what we see,” Jacob says. (There has to be some reality to the regional brand preferences, or Stylefeeder wouldn’t be seeing a doubling in click-through rates when it shows visitors items that are popular in their regions.) “There are 2.5 million monthly unique [visitors] hitting the site, so there’s enough volume that this data matters,” says Jacob. Now the question is—to whom?
Trending on Xconomy
By posting a comment, you agree to our terms and conditions.