eBay Buys Hunch To Improve Long-Tail Shopping Recommendations

Next Story

Ice Cream Sandwich Will Get Flash Support By The End Of The Year

Hunch, a service that provides a “taste graph” of personalized recommendations based on users’ interests, has just been bought by auction site eBay, the companies have confirmed. The amount hasn’t officially been disclosed, but Michael Arrington (who had the scoop this morning) hears that it’s around $80 million.

[Update: We caught up with Dixon and eBay chief technology officer Mark Carges by phone just now, and got some more details on the deal and what it means for both companies. Our notes below.]

Founded in late 2007 and launched in 2009, the New York company will be used by eBay to help improve buying and selling recommendations for its users. From the release:

Hunch’s technology talent and its deep expertise in areas like machine learning, data mining and predictive modeling are expected to help eBay expand and grow merchandising and relevance capabilities to further improve the shopping and selling experience for eBay customers. For example, eBay buyers are expected to benefit from Hunch’s predictive ability to generate meaningful, yet often non-obvious, recommendations for items available on eBay based on their specific tastes.

Cofounder Chris Dixon (a regular contributor here at TechCrunch) says on his company blog that the relationship with eBay started after Hunch began allowing other companies to use its Taste Graph. As part of eBay, Hunch will continue to operate somewhat independently — all of its employees are staying on at its New York headquarters, and the Hunch.com site will stay live.

Hunch had raised around $20 million from investors including Bessemer Venture Partners, General Catalyst and Khosla Ventures and Ron Conway.

Interview notes: 

Dixon and Carges say that the deal will help surface more quality recommendations from eBay’s “long tail” of unstructured listings. Let’s say a coin collector is on eBay looking to add to their collection. As Dixon explains, Hunch might be able to surface relevant items that aren’t obvious, like microscopes that are especially good for coin analysis. Traditional machine learning won’t necessarily be able to identify the same sorts of connections.

Of course, other retail sites, like Amazon, provide recommendations as well — “users who also bought X bought Y” — but those methods rely on existing catalogs, Carges says.

The 20-person Hunch team will begin working with eBay’s data science team “ASAP,” and will anchor the auction company’s physical expansion into New York. Carges is planning to hire more data and engineering employees for that office, along with product-oriented staffers, like designers.

In terms of results for eBay users, there won’t be any drastic changes. Hunch will rather be providing more nuanced recommendations on the back-end, resulting (they hope) in more meaningful discoveries, more stickiness on the site, and ultimately more buying and selling.

The two aren’t commenting on the deal price, or on Hunch’s current revenues. The Hunch.com site will stay live, and will continue to experiment, similar to what eBay has done with local search acquisition Milo last year, according to Carges. There a no changes planned for Hunch’s open API or its other partnerships.

We’re also talking to Dixon, a serial entrepreneur and investor, about having him do one of his TechCrunch Founder Stories interviews with, er, himself… we’ll figure out how to set that up.