turning the messaging service into a targeted-advertising play. BusinessWeek soon named him one of its “top 25 entrepreneurs under 25.” But the messaging service proved difficult to monetize and expensive to run, and Keighran ended up selling his Bluepulse shares and leaving the company in 2007.
He spent the next couple of years as an angel investor and startup advisor, working particularly closely with Aardvark, the real-time question-answering service later acquired by Google. But he never stopped thinking about app search. “After spending time advising and being involved in other people’s visions, I personally was getting really itchy to try and do another company and build something myself,” Keighran says. “One of the things I was really excited about was seeing the shift here in America from consuming services on a Web page to consuming services through an app. That is, to me, as big a shift as going from reading newspapers to reading on a Web browser.”
Keighran located a co-founder in Cathy Edwards, a fellow Australian and machine-learning and natural-language-processing expert who, coincidentally, had once advised Australian wireless giant Telstra that it should buy Bluepulse. Together, Keighran and Edwards decided they would tackle the app search problem head-on. “We believed that if people were changing how they consume things, there was an opportunity to change how they discover things,” says Keighran.
Keighran and Edwards assembled a group of experienced engineers from search companies like Google, Powerset, and Cuil, raised $2.6 million in funding from Blue Run Ventures, ex-Googler Aydin Senkut, and the ubiquitous Ron Conway, and moved into Aardvark’s old headquarters on 10th Street in the South of Market neighborhood. They released the first version of the Chomp app in 2010.
Finding an app on Chomp is a completely different experience from browsing or searching the established app stores, and the difference starts under the hood. Keighran and Edwards wanted their app search service to emulate Google’s search algorithms, which use hundreds of signals about a Web page—e.g., what other pages link to it, the text of the links, and metadata such as title tags—to determine how the page should rank on search result pages. The problem is that apps aren’t like Web pages. There’s no metadata to go on. When a search engine looks at an app’s page in an app store, all it can really see is the app’s name, its category (games, health, productivity, etc.), and the description penned by its developers.
Chomp tries to go deeper by ingesting all of that information—plus all of the reviews left by users, plus blog posts, tweets, and other external data—and using natural-language processing and other techniques to distill it all down to a few keywords that describe the app’s core functions and topics (Chomp’s engineers actually call them “appwords”). “We produce a whole new set of data that describes what an app does, and we use that data for search,” Keighran explains. The bottom line: if you search Chomp for “restaurant guide,” the top three results are—just as you’d expect—Urbanspoon, Yelp, and Zagat.
The other big difference between Chomp and the existing app stores is that Chomp actually feels like a store. (Which the iTunes App Store and the Android Market really don’t, when you think about it. Their search result pages just show screen after screen of little square app icons.) In Chomp, each app in a list of results appears inside its own sideways-scrolling page or “card.” The card shows the name of the app, the price, the percentage of positive reviews, and
