Predictive Networks Testing Biometric Ad Targeting System
The company, based in Cambridge, MA, specializes in tracking and analyzing online behavior to customize ads to individual users. It created Digital Silhouettes, artificial intelligence-based anonymous user profiles, which are the heart of its biometric targeting system.
Biometrics is the science of identifying an individual by measuring physical or behavioral characteristics and comparing them with a library of behaviors or characteristics belonging to many people. Fingerprint identification is probably the oldest and best-known form of biometric identification. Other forms include voice authentication and retinal scanning.
Predictive Networks said last month that it filed a patent application on its biometric technology, which combines network-based clickstream data with biometric data to analyze keystroke and mouse control usage patterns. The system generates an anonymous, random identification number and does not store personally identifiable data in a database.
The company hopes to use the technology as a means to remotely identify household members who use the same computer. Predictive Networks' Internet service provider and cable television clients can use the technology to build profiles of computer users' habits based on data such as which Web sites they visit, how long they stay on each site and which ads they click on. But they cannot tell whether a user who frequently visits a book retailer's site is the same user who also frequents a sports Web site.
With the biometric system, Predictive Networks said, an ISP can build user silhouettes based on distinct usage patterns. The ISP then can infer such information as which household member it is and even the user's sex or age.
The company is working with AT&T's WorldNet Internet service and is serving ads to its low-priced i495 service.
"Advertisers want a way to reach specific users in a household," said Jeanette Medlin, Predictive Networks' vice president of marketing. "The system randomly assigns a digital ID number. We don't know names, addresses or credit card data."
However, biometrics is not without privacy concerns. A number of privacy advocates equate tracking keystroke data and mouse usage to eavesdropping.
Predictive Networks' Medlin, not surprisingly, does not see any privacy concerns inherent in the company's technology.
She noted that clickstream data are analyzed in real time and then discarded, so no personally identifiable data are stored in the company's database.
"Users have control of their Digital Silhouette," Medlin said. "It's always opt-in, and they can easily opt out whenever they want."
Richard Smith, chief technology officer at the Privacy Foundation, Denver, said he does not think online privacy is much of an issue with Predictive Networks' system.
"I see less privacy concerns than normal in this case," Smith said. "What they are trying to figure out is who is using a particular home computer at any one time."
He noted that if the company were using retinal scanning technology, for example, that would be a much larger concern.
"With biometrics, the privacy concern comes in if you can identify one person from among a group of 100,000," Smith said. "I don't see a concern just trying to pick out which person in a particular household is using a PC."
Medlin noted that Predictive Networks has about 1 million Digital Silhouettes in its database. The company hopes to target the system to Internet firms and providers of interactive television service. The system is expected to be available by the fall.
Currently, the system can differentiate among 15 or fewer people on a single device. The company hopes to raise that number significantly in the future.
"We can tell almost instantly who is at the control among a group of 10 to 12 people," Medlin said.