Data Mining – Must-See TV Ads

Americans are rushing to purchase digital personal video recorders like TiVo so that they can zap pesky TV commercials right out of their lives. TiVo sales from August to October alone soared 73 percent compared with the same period in 2002, and the commercial television industry has good reason to fear that devices such as TiVo could spell the demise of an advertising gravy train.

Now advertisers are testing data-mining software designed to get through to TiVo users, according to a recent story in Communications, the monthly journal of the nonprofit Association of Computing Machinery (ACM). But the software that meets Madison Avenue’s needs could create new privacy concerns for consumers because it can profile TV viewers and deliver targeted advertising to them via their beloved TiVos.

This is how it would work: The TiVo would record its patron’s viewing patterns, and software in the recorder would use statistical models?together with any additional information collected from the viewer?to develop a demographic profile. “We already have a large data set from the Nielsen Co. that shows who people are, based on what they watch,” explains William Spangler, assistant professor of IT at Duquesne University and a coauthor of the ACM story. “We know, for instance, that people who tend to watch bass fishing shows weekly and golf and the sitcom Friends are men aged 45 to 50 with a certain income.”

After profiling the viewer, the system would choose specific ads from a stock inventory and download those to the individual’s digital video recorder (DVR). “Our idea is that it would be like being in a movie theater, when ads are presented prior to actual viewing so they don’t intrude on the viewing,” Spangler says. That timing, Spangler hopes, would distinguish TiVo ads from current television ads that are so annoyingly intrusive that they set off the stampede to TiVos in the first place.

Spangler and his colleagues at Duquesne are still developing statistical models that can more precisely predict the demographics of particular viewers. And as they are quick to note in their ACM article, there are prickly privacy issues that must be ironed out before such data-mining techniques can be applied.

Spangler says, for instance, that such a device could work only with an opt-in process?consumers would have to agree to receive advertising in this way in return for some incentives, like getting a better price on their DVR.

That’s a noble thought. But the recent history of online data mining suggests that entertainment industry executives, rather than embracing the idea of opt-in, would inevitably push for an opt-out policy that leaves it to individual consumers to say they don’t want their viewing habits collected in large databases. It could end up like another privacy issue for legislative policy-makers to tackle. The Do Not Call list has been a hit. Can you see a Do Not Advertise registry?