This article was taken from the February 2012 issue of Wired magazine. Be the first to read Wired's articles in print before they're posted online, and get your hands on loads of additional content by subscribing online.
Amazon doesn't recommend a book based on where you read it. Last.fm can't scrobble what mood you're in. But Dressipi aims to do exactly these for fashionistas. Cofounders Sarah McVittie (above, seated) and Donna Kelly (above, far left) have created a context-aware recommendation engine that picks clothes to suit its users' tastes, based on occasion or mood -- an algorithm that knows that you might love a pair of jeans, but not for next week's wedding. "Sarah and I have very different ideas of glamour," says Kelly, 42. "Understanding context and how that maps to someone's personality is key."
To find out a user's "fashion fingerprint", Dressipi first determines their shape -- not by measurements, but by asking them for details of four favourite items sitting in their wardrobe.
Hundreds of features such as neckline and skirt length are identified automatically in each garment. The features are then combined and fed through a set of rules devised by Dressipi's six stylists to build up a probabilistic model of what a user will like, shown as a points score. "We're on our fifth iteration of style, and it's currently around 80 to 85 per cent accurate," says McVittie, 34.
The site came out of beta in November and attracted 20,000 users by the end of its first week. And the west London-based company is, at the time of writing, about to launch an iPhone app, so "you can have a portable fashion fingerprint," says North. "The aim is, you walk into a retailer, scan a barcode and the app tells you the dress will go well with that jacket you bought two weeks ago." The millions of data points that Dressipi collates could offer "a far better experience" than human stylists. The machines may or may not take over the world, but they will ensure we're impeccably dressed.
Dresses and location were provided by Wish Want Wear.
This article was originally published by WIRED UK
