Meet Ness, the Pandora for restaurant recommendations

Even in a city chock full of great restaurants, it can be really tough to find a new one I’ll be sure to like. I’ve found that the more that sites such as Yelp or TripAdvisor get packed with reviews, the longer it can take for me to pick out a new place to try. Say a restaurant has a ton of five-star ratings from Yelp users: What if those people loved the restaurant’s loud pop music, or the long wait for a table? Lots of folks think those are the hallmarks of a hot restaurant — but not me.

That’s where a startup called Ness Computing comes in. On Thursday, Ness launched its debut iOS app, a “personal search engine” that aims to provide highly personalized restaurant recommendations. In an interview this week, Ness co-founders Corey Reese and Paul Twohey took me through an extensive demo of the app, and I have to say, it’s pretty awesome. Essentially it’s like a Pandora Radio for restaurants: From the start it’s almost creepily accurate with its recommendations, and the more you use it, the better it gets.

Tailored restaurant ratings

Here’s how it works: You sign into Ness using either your email address or Facebook credentials. If you sign in with an email address, it asks you to begin rating nearby restaurants that you may have visited to start getting an idea for the kinds of places you like (it knows your location via your mobile geo-data, but you can enter in any place in the United States too.) Once you have rated ten places, it can begin to start offering recommendations of other spots you may also enjoy. Ness assigns each restaurant with a so-called “Likeness Score” of between zero to 100 that predicts how much you’ll like it.

But if you connect Ness with your Facebook or Foursquare account, you’re on the app’s fast track. Sifting through all the data provided by these services, Ness automatically aggregates all your friends’ recommendations and check-ins as well as your own to estimate what kinds of places you might like. It starts assigning Likeness Scores to restaurants right away, and you can train the app to better understand your tastes by adding your own ratings to restaurants you’ve already visited.

Behind the slick user interface, serious tech muscle

Ness is the product of 18 months of work from a 15-person team based in Los Altos, Calif., south of San Francisco. Backed with $ 5 million from a star-studded group of venture capitalists including Khosla Ventures, Alsop Louie Partners and Eric Schmidt’s Tomorrow Ventures, the company has made some pretty hardcore technology under the hood to do what it does. Ness’ staff is comprised of former engineers at Apple, Google, Ning, and Palantir with specialties in applied machine learning, natural language processing and more. The company leases its own servers to process the large amounts of data it sifts through to determine Likeness Scores, and Ness has already filed a handful of patents for its search technology.

Ness plans to eventually monetize its app, which is free, with targeted advertising — but the company’s co-founders tell me that’s a bit further out on the horizon. “Right now, our revenue is user happiness,” Twohey said. “Our model will never be to give all the data we collect to advertisers. We realize that it’s a trust-based relationship we have with our users.” Users could be hugely turned off if Ness starts placing paid results alongside its real recommendations, so the company has to be really careful about how it chooses to monetize when it starts needing to pay the bills.

It’s not just for foodies

And Ness’ co-founders say that restaurants are only the beginning. The company’s core technology could be used in the future to find upcoming concerts you would probably like, or nightlife spots that would suit your tastes, Reese said. “In the same way that Google’s objective is to organize the world’s information, at Ness our goal is organizing the world’s opportunities.” It’s an ambitious aim, to be sure, but it’s also a worthwhile one.

Here are some screenshots of Ness at work (click to enlarge):