WWD.com/media-news/digital/parsing-personal-style-to-push-new-looks-462726/
government-trade
government-trade

Parsing Personal Style to Push New Looks

StyleFeeder lets users share favorite products and automatically recommends new ones based on their taste.

What do 14-year-old boys who wear Quiksilver like to drink? Sprite. And 15-year-old girls, it turns out, are wild for Nutella.

This story first appeared in the March 5, 2008 issue of WWD.  Subscribe Today.

Philip Jacob, StyleFeeder founder and chief technology officer, knows this because his Web site tells him so. StyleFeeder, which he created in 2005 and has now grown to eight employees, lets users share favorite products and automatically recommends new ones based on their taste.

It’s a new spin on so-called social shopping, or product bookmarking, sites that seems to be taking off. The site gets a self-reported half-million visitors a month, and in January received $2 million in venture funding led by Highland Capital Partners of Boston (bringing total investment to $3 million).

Similar sites include Style-hive, Kaboodle, Wists and ThisNext.

The basic idea is that users bookmark favorite products online and can look at the products others have bookmarked. Items can be searched by tags, or keywords, as well as by the person who has bookmarked them. If you like someone’s taste, it’s easy to follow along as she makes new discoveries. Most of the sites also usually include personal profiles, blogs, search, ratings and other features.

What makes StyleFeeder unusual is that it also uses a technology known as collaborative filtering to automatically recommend items to users. The software creates a giant matrix or spreadsheet of items and their relationships. Then it looks for patterns among users who rate items similarly.

The basic idea is that if two people like product X and one of them also likes product Y, then they probably both like Y. Amazon.com and music site Last.fm also use collaborative filtering to recommend new books and music.

When a user registers with StyleFeeder, he or she is asked to rate four or five items right away. “We use what we call polarizing items,” said Jacob, meaning controversial styles that are typically loved or hated. “The Pillsbury Doughboy sweater is the most hated product on our site,” he said, referring to a navy sweater sporting the Pillsbury Doughboy image. (The item can no longer be searched on the site, but the sweater remains listed on carebearlove’s page as a favorite.)

The iPod and iPhone in all their variations are the most loved products on the site. So knowing a user likes the iPhone “is totally useless to us,” said Jacob. “That just means you’re like everybody else.”

The site takes into account other pieces of information to make recommendations, such as a user’s age and gender, as well as the words associated with items the user likes, and how active the user is on the site.

While the resulting data could be sliced and diced in an infinite number of ways, there are a few obvious taste clusters that show up again and again. The 14-year-old boys, for one, also like snowboarding, the Wii game console, Adidas shoes and music by My Chemical Romance. Women who have recently graduated from college prefer Kate Spade, certain brands of makeup and “lots of expensive handbags.” Bookworms tend to import Amazon wish lists of 5,000 books and little else. There is also a

prototypical “sporty” male type. And “we’ve got our fair share of the metrosexual crowd,” said Jacob. This includes men in their late 20s and early 30s who like modern furniture, designer clothing and electronica. Brands include John Varvatos and Giorgio Armani. If they tag any books, it will probably be business and marketing nonfiction. “You tend to see a lot of products from Bluefly in this kind of crowd.”

The number-one user of the site has rated 80,000 products, and users must rate more than 30,000 to get into the top 10. StyleFeeder is using Glam to sell advertising, and it also collects affiliate fees when someone buys a product recommended on its site. Jacob said he hopes the company will be cash-flow positive by the first quarter of next year.

The site has grown significantly in the last six months, thanks to a Facebook application, a marketing partnership with Shopping.com and interest around products from celebrities such as the Olsen twins.

Mary-Kate and Ashley Olsen created a StyleFeed of products they like and sell, and linked to it from their own Web site. As a result, fans came into StyleFeeder and created their own StyleFeeds, said Dina Pradel, StyleFeeder vice president of marketing. Fans on Facebook and MySpace also added products the Olsen twins sell to their pages. “It creates viral marketing and sales,” she said.

Over the summer, the company launched its Facebook application, which lets users post a feed of their favorite items to their Facebook page. As of press time, it had more than 10,000 active daily users.

Handbag designer Lori Naon saw an increase in sales and traffic on her Web site after she put some styles up on StyleFeeder last year. She also writes a blog on StyleFeeder about accessories from independent designers. “We can’t afford a big p.r. firm to get the word out and we rely on the Internet,” said Naon, a former textile mill representative based in Bainbridge Island, Wash. Her company is Hardware Handbags and its designs are manufactured locally in limited quantities.

“StyleFeeder marries the interactivity and serendipitous discovery you would find in an offline mall experience with the immediacy of the Internet,” said Gaurav Tewari, Highland Capital senior associate. “If you look at what the system does, it learns the user’s sense of style over time, and lets users discover products they’re likely to love but unlikely to run across if they used a portal or search comparison shopping engine.”

A visitor can learn a lot about someone by looking at her StyleFeed, said Jacob. “How we choose to express ourselves, display ourselves at home and in the types of intellectual and cultural activities we are drawn to come shining through in the things we buy and are interested in,” he said. “If you take five friends and remove their names and look at the products, it’s very obvious who they belong to. People’s preferences come shining through.”