Polyvore creates 'style graph' to help you buy more clothes
The image-heavy social media site rolls out a new iOS app that attempts to do what Amazon can't: personalized fashion recommendations for each of its users.
Polyvore is trying to go for the "holy grail" of fashion: knowing exactly what kind of clothes you would like to buy.
The fashion-centric social media site released a new iOS app Tuesday that includes a new algorithm to suss out your tastes in clothes, shoes and everything in between. An updated Android app is also on its way.
CEO Jess Lee said searching for typical products, the main signals needed are more straight-forward, like price, color or material. "But, fashion isn't like that," she said. "You really need a ton of data in real-time."
Personalized fashion recommendations means accounting for seasonal styles or constantly-changing style trends. For Polyvore, a network that is part book-marking tool and part e-commerce site, this means harvesting the data collected from user activity and marrying it with the work of its editorial team, which focus on what's trending. The site's users curate fashion items and style them onto board to create what Polyvore calls "sets." Each item used in a set links back to the site of the retailer who sells the item, and Polyvore gets a cut of the transaction.
Polyvore, which sees 20 million unique visitors a month, says that 3 million sets are created each month. Lee said this is the kind of data large e-commerce sites just don't have. Large e-commerce sites like Amazon and eBay have been ramping up their fashion efforts in recent years -- through larger, detailed images for listings and promotional stunts like participating in New York Fashion Week -- in hopes of breaking into the world of loyal fashion shoppers. But, it's a hard nut to crack.
"Amazon has amazing transactional data, we have amazing style data," she said. "It's not just about what you buy. It's what you want to buy."
The new search algorithm creates something Polyvore is calling a "style graph." A nod to Facebook's "social graph" concept where a user's activity shapes what they actually want to see on their feeds, the style graph takes into account a person's clicks, likes and sets created on Polyvore. It also takes into account what similar people have also liked and what other items would go well with those products.
To make this work, Polyvore also needs more user input. The more a user feeds into the system, the better the algorithm gets at predicting what a user wants to see, Lee said. The mobile app includes new ways to like and discard things, making it easier for people to add to the algorithm. Users can double tap an item to like it, and press and hold to make an item disappear from their feed.
If it works, Polyvore will have figured out a way to surface exactly what a women wants to wear, and ultimately, what she may buy. And that's good news to the retailers who are linked to the site.
"Personalization is the holy grail," Lee said.
Update, 10:03 a.m. PT: Adds number of Polyvore users.