New Study Casts Doubt on Web Advertising Data

The quest for reliable data on Web advertising is a priority for marketers, as the medium has started to attract real money: An estimated $1 billion was spent on Internet advertising last year. That is expected to double by 2000, reflecting the Web’s growth in traffic.

Over the last four years, at least half a dozen major firms have sprung up to provide audience measurement services for the Internet. They gather data using customized software either from the “server,” or Web site side, or from the consumer side, via representative samples of Web surfers.

Server-based firms such as Internet Profiles (I/Pro) and Andromedia sell and support software on a client’s Web site that tracks visitors to the site. Panel-based firms such as Relevant Knowledge recruit thousands of Internet users to make up a panel representative of Internet users in general; the firms then make projections based on that sample, much as the Nielsen TV ratings system does.

*

These companies all have their own claims to superiority: Server-side trackers say they can provide hard data on unique customers and “click-throughs” (the number of people who click on an ad to get more information); panel-based firms aim to give a more detailed demographic picture of who is going where on the Internet.

But according to a newly published study from two USC professors and the former chairman of Web-tracking firm Netcount (now owned by I/Pro), none of these firms is yet giving a reliably accurate picture of how many people see an online ad and who those people are.

Instead of the totally quantifiable medium promised in the Net’s early days, says co-author Xavier Dreze, statistics vary wildly. “The assumption has been that on the Internet, you could measure everything accurately. That’s not the case now,” he said.

One common problem, according to co-author Fred Zufryden, is that server-side methods usually can’t track individual behavior, since users may be assigned a different identity by their Internet service providers every time they sign on. “So it’s almost impossible to tell whether one person requests a page five times or five people request a page one time,” Zufryden said.

*

Some Web sites employ “cookies,” or registration processes to try to get around this problem. But, said Zufryden, both methods are easy to circumvent: A user can eliminate the cookies file from his or her computer, thereby erasing that trail, and Web site visitors can easily lie about their identity and characteristics when queried.

Executives at Web measurement firms maintain that while their services may not be 100% accurate, they are at least as reliable as ratings for mass media such as TV and radio.

“TV and radio aren’t really measurable in any concrete way,” said Kent Godfrey, chief executive of San Francisco-based Andromedia, a leading supplier of server-side research tools.

“The Internet is infinitely more measurable than television, which is tracked through representative samples [of audiences]. So saying that Internet figures are unreliable is all relative,” Godfrey said.

*

Jeff Levy, chief executive of Relevant Knowledge, a panel-based measurement firm in Atlanta, agrees that methods for Internet audience measurement are still evolving.

“What gives many marketers pause is the disparity between the different kinds of measurement companies,” Levy said.

“There are huge differences between data from server-based and panel-based [companies]. Unfortunately, that’s partly just due to the fact that the real world doesn’t equal the observable world. These differences need to be explained better and quantified more rigorously to make marketers feel comfortable. Right now, they can think, ‘Hey, they’re trying to pull a fast one on me.’ ”

Another way to supply a comfort level to advertisers is to have figures audited by a third party. I/Pro provides auditing services, in addition to two companies established in the auditing of print materials: the Audit Bureau of Circulations and BPA International.

*

Evelyn Hepner, vice president of sales and marketing for the audit bureau’s ABC Interactive subsidiary, said advertisers looking for greater assurance have driven a boom in Internet auditing over the last two years.

“Agencies and advertisers are familiar with numbers that are audited,” Hepner said, “and this is a medium, like print, that is census-based. It can be counted, rather than projected.”

Apparently, not counted accurately enough, according to this study and those in the measurement business. But for now, marketers who want the Internet to be part of their advertising strategy are gathering what information they can, usually through a combination of server-based and panel-based research.

“I think the study has made a strong case for auditing,” Hepner said.

(BEGIN TEXT OF INFOBOX / INFOGRAPHIC)

Tracking Eyeballs

Internet audience measurement services fall into two broad categories: server-based, which uses software to track data about visits to a specific Web site, and panel-based, in which companies install tracking software in the personal computers of Internet users and use the information to make projections based on that sampling.

Here are some of the major audience measurement companies:

Server-based

I/Pro

Andromedia

Web Trends

Panel-based

Relevant Knowledge

Media Metrix

Net Ratings

Source: The companies