Net Ad Measurement Is Inexact Art

Welcome Guest. | Log In| Register | Membership Benefits

August 06, 1998 (3:13 PM EDT)

By Mo Krochmal,

Measuring who reads Web advertising is not a hit, or even close to it, according to a study conducted by researchers at the Marshall School of Business at the University of Southern California.

According to the study, published in the current issue of the Journal of Advertising Research, the technology used to measure how many people see an ad can be highly inaccurate.

Error rates for sites that track users through a log file of IP addresses are as great as 30 percent, said one of the co-authors of the study.

"What we asked was if this was a real problem or just an annoyance," said Xavier Dreze, an assistant professor of marketing at the university. "What we found out -- if you use those log files, you get errors of 20 to 30 percent," he said. "If you are underestimating by 20 or 30 percent, that is a big deal and it's something that needs to be addressed."

Accurately measuring traffic to a website is a critical issue as the Internet, fueled by growing advertising revenue, evolves into a mass medium. Last year, Internet advertising revenue grew to $906.5 million.

The researchers tabulated traffic on five major websites. They used IP addresses to infer distinct users, and then compared that to information gathered from data obtained from the same sites.

This time around, however, mandatory registration enforced viewer restrictions. The authors said using IP addresses as a means of identifying website "hits" led to a 39 percent underestimation of visits -- a 64 percent overestimation in page views and a 79 percent overestimation in the time spent on each visit.

The science of measuring viewership of Internet advertising is still being developed. Originally, hits were the yardstick to measure Web viewership.

But page views, which is an aggregation of hits, has become a more accepted measure of traffic and viewership. Now that measure has also been clouded, because measurement software cannot count pages that are served from caches.

A whole segment of technology is devoted to producing more sophisticated metrics, said Drew Ianni, online advertising analyst for Jupiter Communications in New York, a consumer online research company.

"Overall, the industry has moved well beyond the discussion of hits, cached pages, and counting log files, " he said.

"All you can really do is go off the publisher's statement of the people a site is reaching and the frequency. You have no choice but to believe him."

In June, the Coalition for Advertising Supported Information and Entertainment offered a list of proposals for global standardization of measurements.

The Internet Advertising Bureau is also preparing to develop standards for audience measurement.

Such standards are 12 months to 24 months away, said Ianni.

"There has to be greater consistency from advertisers on what they want," he said. "Until evolving models settle down and are fleshed out, it will be difficult to get a measurement standard."