have different thresholds for what’s appropriate.” Disney, for example, probably wouldn’t want to be seen on a page containing a story about alcoholic beverages. Budweiser, on the other hand, might covet that spot.
Peer39’s semantic technology can also rapidly detect changes in the sentiment behind certain words. For example, prior to the recent Missouri tornadoes, the term “Joplin” only appeared on .16 percent of pages, and 95 percent of those were music sites related to the legendary Janis Joplin. After the tornadoes, “Joplin” showed up on 2 percent of all Web pages, more than 90 percent of which Peer39 identified as part of a category it calls “Death/Disasters”—a negative environment for ad placements in Peer39’s vernacular. That made it easier for advertisers averse to negative content to steer clear of the disaster coverage.
Ellenthal says the secret to Peer39’s platform rests in its ability to use both natural language processing and machine learning. “They balance each other out,” he says. “Natural language processing looks at relationships between words. It’s very accurate, but not very fast. Machine learning understands how content changes over time, and it’s very fast.”
Peer39’s customers are ad networks and exchanges, agencies, and publishers. But Ellenthal believes the company’s biggest growth opportunity lies in real-time ad bidding platforms such as AdMeld, which was recently bought by Google (NASDAQ: [[ticker:GOOG]]) for $400 million. AdMeld is one of Peer39’s customers. “They use our service to provide buyers with more information about quality, safety, and the topic of pages,” Ellenthal says.
Canaan’s Lee adds that the popularity of real-time ad bidding has exploded in the last year, creating an opportunity for all ad-tech companies. But what he particularly likes about Peer39’s platform is that it’s both fast and scalable. “A lot of companies have developed technologies that are useful on a small scale, but when you get to a billion impressions a day, they break,” he says.
Ellenthal says that scaling up Peer39’s platform to meet the demand from AdMeld and other real-time players hasn’t been difficult. In January of 2010, Peer39’s technology was processing about 30 million page impressions a month, he says. “Right now we’re looking at 3 billion a day,” he says. “We’ll be close to 10 billion by the end of this month.”
Other ad-tech companies are starting to raise the interest of the venture community. On June 9, New York-based Taykey—another company founded in Israel—raised a $9 million Series B, led by Sequoia Capital, Softbank Capital, and Crescent Point. Co-founder Amit Avner says Taykey’s technology is also non-cookie-based, but that’s where its similarities to Peer39 end. Taykey’s algorithm uses real-time data and trend analysis to predict where an advertiser’s audience is likely to go next on the Web. The platform is most useful for companies that want to advertise on social-media sites or search engines, Avner says.
In a beta test, Pepsi used Taykey for two Facebook campaigns. Pepsi’s goal was to get 20,000 “likes,” and it ended up with 46,000, Avner says. One of Taykey’s goals is to improve the technology so the company can also participate in the real-time ad bidding boom, Avner says.
Ellenthal is encouraged by the venture community’s support for companies in his space, but he knows that Peer39 will have to do plenty of evangelizing to get stalwart brands to embrace newfangled ad technologies. “For us to continue to grow we’ll need to see more brand dollars flowing into these exchanges, and much bigger campaigns,” he says. “We need to make those brands comfortable leveraging these new technologies. They need to know their ads won’t end up on pages that are inappropriate, but rather in environments that are relevant to what they’re trying to do.”
