StyleFeeder—Facebook’s Leading Shopping Engine—Thinks Big with Small Series A Round

While Amazon and eBay may dominate shopping in most of the Web universe, things are still up for grabs on planet Facebook. There, the leading shopping application is from a two-year-old Cambridge, MA, startup called StyleFeeder. More than half a million people have made StyleFeeder’s shopping app part of their Facebook profiles—which is about 50 times as large as the Facebook user base claimed by the nearest competitor, eBay.

The personal shopping engine at StyleFeeder’s own website boasts a similar number of users. And this week the startup literally capitalized on its lead, raising $2 million in Series A financing from local venture firms Highland Capital Partners and Schooner Capital (the same two firms that provided the company with a $1 million seed round last year).

“Obviously, at some point, an eBay or an Amazon will do something innovative with their Facebook apps, but we’ve established such a big lead right now that we want to keep being pervasive,” says Shergul Arshad, StyleFeeder’s vice president of business development.

I spoke with Arshad and Phil Jacob, StyleFeeder’s founder and CTO, by phone yesterday. They said the success of the StyleFeeder Facebook app, launched last June, was something of a surprise, even to them. “The Facebook platform was very new at that point, and I don’t know if anyone really knew what it was going to become, but it looked pretty interesting,” says Jacob. “We looked at the API [the application programming interface, the specifications outside programmers can use to interact build software that interacts with Facebook’s database] and it looked like we would be able to weave ourselves into the social fabric of Facebook pretty easily, in a way that was truly useful.”

And a useful Facebook app is not a common species, as you may know if you’ve been anywhere near the site in the last year. Puerile gag applications like “Zombies” and “Vampires” have proliferated at viral speed, only to shrivel as the novelty wears off. (“Facebook apps are for toddlers,” the Wall Street Journal‘s Kara Swisher wrote last October—and out of everything she’s said about the company and its founder CEO, Mark Zuckerberg, that’s about the nicest.) Stylefeeder, by contrast, gives readers fairly accurate (in my experience) recommendations about products they might want to buy, based on the preferences of people with similar tastes—and the more products users click on, the better StyleFeeder gets at matching them with their “Style Twins”.

“There’s kind of a backlash today against a lot of these useless Facebook applications,” says Jacob. “The utility we provide has been a big factor” in StyleFeeder’s relative success on Facebook, he believes.

StyleFeeder is by no means the first Web company to offer product recommendations based on its understanding of users’ tastes. So-called collaborative filtering software goes back nearly to the beginnings of the Web and is a key element of venerable e-retail sites like Amazon. But StyleFeeder’s success on Facebook isn’t the only thing that sets it apart from the pack; the company has introduced a couple of twists on collaborative filtering that are helping to keep the technology relevant in the Web 2.0 era.

For one thing, rather than limiting its style recommendations to products culled from a single catalog, the way most e-retailers do, StyleFeeder allows users to rate and recommend any product found anywhere on the Web. Jacob says that started out as a practical strategy: the company wanted to let users decide what to share. But in any case, populating the StyleFeeder database with products from just a few vendors would have made it “incredibly narrow,” in Jacob’s words. “People want choice and depth, and once you’ve experienced this very, very long tail, anything less starts to seem underwhelming.” (If you think Amazon has a lot of products, get ready for another multiple: there’s 20 times as much stuff in StyleFeeder’s database.)

Stylefeeder on FacebookIn addition, StyleFeeder’s brand of collaborative filtering hinges on a unique implementation of a machine-learning algorithm called maximum margin matrix factorization, or M3F, which—naturally enough, this being Cambridge—hails from the MIT Computer Science and Artificial Intelligence Laboratory. Jason Rennie, who received his doctorate from CSAIL in January 2007, is StyleFeeder’s head of machine learning technologies; he’s tuned M3F in a way that helps StyleFeeder determine exactly which attributes of each user—such as their gender, age, location, or previously expressed preferences—seem to make the most difference in their product choices, and it can apply those insights to produce more accurate recommendations for similar users. M3F “performs several percentage points better than the state of the art” in collaborative filtering, Jacob says. (Which makes it the new state of the art, the way I read it, but I won’t quibble.)

StyleFeeder is using the venture infusion to hire more developers—which Jacob says hasn’t been a problem so far, despite the Boston area’s reputation as an unfavorable environment for consumer-facing Internet startups. The company also hopes to acquire more users over the coming year and “work toward profitability,” according to Jacob. Highland Capital and Schooner Capital “are very big believers in what we’re doing, as evidenced by the fact that we didn’t go outside our original funders for this round,” says Jacob. “And frankly they are incredibly well connected and add a ton of value to what we do.”

StyleFeeder’s $3 million in funding to date isn’t a lot, especially compared to figures like the $10 million raised by Needham, MA-based Matchmine, another company focused on facilitating online shopping by determining Web user’s personal preferences. But Jacob says it’s plenty for now. “More and more, people are losing sight of the fact that in the Web 2.0 era you can build a large, consumer-facing website—if you have a good product and you can communicate it well—on much smaller amounts of capital,” he says. “We’ve taken in only a million so far, plus the new $2 million round, and we’ve already got metrics like half a million users. All of our infrastructure technology stack is open-source, our technology costs are really low, and we are not spending money on Superbowl ads and this kind of crazy stuff.” And that, in the end, may be the real way to overtake Amazon and eBay.

Wade Roush is a freelance science and technology journalist and the producer and host of the podcast Soonish. Follow @soonishpodcast

Trending on Xconomy

By posting a comment, you agree to our terms and conditions.

2 responses to “StyleFeeder—Facebook’s Leading Shopping Engine—Thinks Big with Small Series A Round”

  1. Fabio says:

    Let’s start with a disclaimer, my company offers a service which is for many aspects similar to StyleFeeder (we’re based in London, England though). However I really like what these guys are doing and I’m a bit envious for a few things they have! :)
    But there are at least 2 things in this article which are really too far-fetched. The first is this statement: “If you think Amazon has a lot of products, get ready for another multiple: there’s 20 times as much stuff in StyleFeeder’s database”. C’mon, don’t be silly. I wouldn’t believe it even if I see it, it’s just non sense. Period.
    The second thing is this claim: “M3F (maximum margin matrix factorization) performs several percentage points better than the state of the art in collaborative filtering”. Well, this could be absolutely true but I find it pretty hard (at least) to standardize collaborative filtering and therefore trying to benchmark it or claiming to be several points ahead. Saying “we’re the best” would be more convincing honestly. Also, with all the different applications for collaborative filtering ‘social shopping’ is definitely not the most popular and therefore not particularly prone to benchmarks and rankings.

    You might think I’m a hater, but I’m not. I’m just trying to be reasonable and honestly I’m fed up with the PR hype which sees readers as a gullible audience who swallows anything.