Software tracks Web habits, click by click

Predictive Networks, a start-up based in Cambridge, Mass., is using artificial intelligence to map out the kind of person a consumer is and the kinds of ads he or she might want to see. Investors seem to like the idea--the yearling company plans to announce today that it has raised $45 million in a second round of funding led by Battery Ventures and Advent International.

The company's software, used by several Internet service providers including PSINet and AT&T WorldNet, gathers data about consumers based on the sites they visit and then delivers targeted advertising to those visitors. For example, if someone travels to both The New York Times' Web site and a women's Web site, an ad might pop up from Condé Nast Publications' House & Garden, a high-end interior design magazine.

Devin Hosea, the company's 30-year-old founder, said this process is done with complete anonymity. "Every user starts out as an anonymous dot on a piece of paper," he said. "But as you visit sites, we may see that you're slightly more affluent or slightly older than the average Internet user."

All the "clickstream" data, such as which sites a person visited, is thrown away, Hosea said.

Predictive's winning of a relatively large funding round in the current dot-com drought shows a continuing appetite for targeted advertising services. As clickthrough rates--the rates at which Web surfers actually click on an ad--plummet online, targeted advertising is the proposed remedy, despite swirling privacy worries about the means to deliver those ads.

To address worries about what happens to consumers' personal information online, a flurry of companies have come on the scene to offer anonymous software that keeps a lock on email, credit card numbers and shipping addresses.

"Privacy and security is looming as the Achilles' heel of e-commerce. And what Predictive is trying to do is let consumers shop without giving up their privacy," said Peter Bernstein, owner of New Jersey consulting firm Infonautics Consulting. "This is permission-based commerce."

Bernstein sees Predictive's software as a boon for ISPs, advertisers and consumers. The theory is this: ISPs using Predictive's software can charge more for advertising because it is targeted; advertisers reach an audience more likely to be interested in their products; and consumers can sign up for free or less expensive Internet connections while generally maintaining anonymity.

The company, which shares ad revenues with the ISPs, garners ad clickthrough rates as high as 6 percent to 10 percent, Hosea said. The industry average hangs well below 1 percent.

Internet service provider IDT is using the software with its free ISP, FreeAtLast.com. AT&T, which has a $5 million contract with Predictive, uses the software with its $4.95-a-month subscriber service.

Predictive, which said it has raised about $55 million since its inception, plans to use the additional money to expand relationships with wireless and cable companies and to grow internationally. The company has talked to TV program-guide company Gemstar-TV Guide International, Hosea said, in the hopes of offering an interactive TV guide feature that would show viewers only programs or movies they are likely to watch based on previous watching habits.

Smaller investors in today's round include investment firm C. E. Unterberg Towbin, Japanese carrier NTT, telephony company Net2Phone, market researcher NetRatings and PSINet.