Predictive Predicts Major Success
Page 1 of 2
When Devin Hosea started Predictive Networks nearly two years ago, he wanted to take advantage of the Internet's promise as the "Holy Grail" of advertising and marketing, able to target all consumers who used it and personalize every ad they encountered.
But, building on his background in artificial intelligence studies both as a Princeton undergraduate and as a National Science Foundation fellow (as well as experience on Wall Street and as a venture funder with GE Capital), Hosea had a technological vision different from either online ad networks like Engage or DoubleClick, or ad-driven, subscription-free Internet service providers.
Hosea also believes the vision solves the privacy problems generated by online tracking.
While some have criticized Predictive's technology, so far Hosea's vision has enabled the company which unveiled its service in the spring, to sign up channel partners such as AT&T, IDT and PSINet and create a base of 500,000 people using the Cambridge firm's software to receive targeted online ads.
Doug Kingsley, a managing director in Advent's Boston office, said, "Quite simply, this company has the ability to change the way advertising is conducted on the Internet and across different types of networks, including wireless and cable television."
So what's different about Predictive's technology? Rather than relying on cookies (files planted on user hard drives when users visit individual websites) or creating user data files that reside on an ad network's or local ISP's servers, the software tracks users and collects data from the network connection points of major carriers.
Having software that interacts inside the network, rather than with individual computers or servers on the Internet's "edge," means that Predictive has to partner with large network players. It also means that the software can track not just Web users but cable television watchers or those using wireless connections. (Hosea says the firm will use more than half its $45 million windfall to attract cable and wireless partners.) The Predictive system tracks visited websites and, using mathematical models and without need of a user's "real" identity, creates a "digital silhouette" based on more than 100 criteria, such as likely age, income, location and interests. The more websites visited, the more the system "knows" about each user and the better it can predict the kind of messages she wants to see. Since networks using Predictive software are likely to have users click through to more ads, they can charge more to advertisers and less to subscribers. (AT&T, for instance, charges $4.95 for its monthly service that uses Predictive software.). Predictive's revenue model is based on dividing this extra income with networks. On the privacy issue, Hosea says Predictive addresses the problem in three ways. First, all users must "opt in" to a service using Predictive software. The company requires channel partners to sign up subscribers only upon receiving "express and informed consent" -- that is, approval of obvious, easy-to-read langu