Former 24/7 Real Media Exec Jack Smith Replaces Bossy Social Media Ads with Technology that Really Listens
If there was ever any doubt that search engines could learn to speak English more fluently, here’s a developer whose program can tell us in complete sentences how it works. Former WPP and 24/7 Real Media executive Jack Smith met us in Brooklyn to talk about his new venture, Solariat, an analysis system that monitors entire conversations rather than using geolocation and key words to suggest similar threads. With $1 million in seed funding from KPG Ventures, the product launched on August 6th after Beta testing with the company’s first customer, consumer website About.com.
The partners at KPG Ventures had been interested in working with Smith even before the product was conceived, Smith said. About a year ago the venture capitalists introduced him to his business partner, Jeffrey Davitz, who founded the Social Computing Group at SRI International’s Artificial Intelligence Center. Davitz already had experience improving conversations within the U.S. military’s private social networks. “Within two weeks we figured out what we were interested in doing,” Smith said.
They started dropping into online forums to see how search engines worked in the context of social media. “When I looked at the sites I thought there was something missing, but I wasn’t quite sure what it was,” Smith said. While on his Facebook page, he noticed that the analysis tool that used his profile information, such as age and location, to generate ads hadn’t exactly captured his essence. A banner ad for car insurance that read, “’44 and driving?’ didn’t relate to anything on my screen,” he explained.
The plan was to be more responsive and less invasive. “People on social media are looking for information,” Smith said. Solariat wouldn’t use geolocation to target consumers, nor would it try to collect personal information to determine its market. “We don’t want to know who says what,” Smith explained, “we only want to know what’s being said.”
In order to create a phrase book for the system to use, Smith would have to find consistency within an inconsistent forum. “Editors follow a style sheet,” he said. Most people conversing on the Web do not. But Smith noticed that a lot of online conversations have similar styles because online communities work within the same constraints: there are multiple people talking at once, the conversations are not always in real time, and sometimes, as with Twitter, there are character limits. Using computer analysis and human testers for quality assurance, Smith and Davitz came up with a response system that’s accurate enough to know that “RU” means “are you” and that has a click-through-rate of 10%.
“We’re not pretending to be human,” said Smith. In the words of one of his clients, he said, “‘ it’s the next best thing to having a human assistant around to monitor conversations on your behalf.’” Solariat might not be witty enough for the Algonquin Round Table, but it earns points for handling passive-aggressive people who write, “I have a headache” and then wait for a response rather than just asking for some pain killers. In the screen shot (below) Solariat generates some suggested links from a real conversation between About.com columnist Mary Hartley and one of her readers.
According to Smith, having About.com for a client was a “huge coup” for the company because the site has stringent requirements. Smith, a former Brooklynite, commutes between Solariat’s two offices in San Mateo, CA and New York City and plans to take the conversation analysis system to other businesses. “We’ve spent 99% of our time building and no time selling,” he said. “We wanted to make sure it worked.”
Launch a social media campaign that will build your brand and deliver results in our online 
FishbowlNY Twitter feed loading...
