But jaws dropped when February figures were reported March 19. AltaVista's expected increase in traffic failed to materialize; Media Metrix recorded a drop in visitors to the site - despite the inclusion of Raging Bull's 593,000 visitors. The decline to 12.3 million visitors pushed AltaVista down the rankings to No. 13.

The timing couldn't have been worse. The next monthly traffic report would not be available before AltaVista's IPO, which was slated to price April 17. AltaVista's bankers postponed the offering following the Nasdaq's dive over the week of April 10. Though the stock-market drop was the big factor in AltaVista's pulled IPO, journalists also pinned blame on the dramatic drop in February's traffic.

What was particularly galling to AltaVista executives was that they had reason to question Media Metrix's numbers. The portal's internal records showed that visitor totals climbed 7 percent from January to February. "The inaccuracy, from our perspective, was devastating, and we're still trying to recover from it," says Ralph DiMuccio, former research manager and now manager of industry relations at AltaVista.

Just as a Nielsen rating can make or break a television program, Media Metrix ratings and those of its competitors - Nielsen NetRatings, PC Data and recent entrant Net Value - have become gospel in the Internet Economy. Web businesses live and die by their traffic figures. It wasn't designed to be that way, however. The ratings were originally intended to provide online advertisers with an independent, third-party opinion on site audience size and quality. However, the use of these numbers has expanded far beyond the walls of ad agencies to the trading floors of Wall Street.

"These ratings [are] the coin of the realm by which business development, advertising, sponsorship deals and, ultimately, these companies themselves are valued upon," says Rich LeFurgy, president of the Internet Advertising Bureau in New York and general partner at Walden VC. Five independent university studies showing that ratings data move Net stocks or influence company valuations back up LeFurgy's point.

Numbers that carry such weight should be reliable, and ratings firms stand by their accuracy. "The importance and accuracy of our numbers is the most important thing we do here," says Tim Meadows, executive VP of products and services for NetRatings. "Whether it is a billion-dollar stock trade or a million-dollar ad buy, we feel it is our responsibility to provide the highest quality data."

But many industry insiders and outside experts say ratings figures are often flawed and wildly inconsistent. One study showed that ratings firms sometimes overestimate traffic by 300 percent in comparison to the number of visitors recorded on a site's own internal servers. Smaller sites, and those that get much of their traffic from workplace surfers, are especially vulnerable to miscounts.

Although sites have long worked with ratings companies to try to improve the data, little progress has been made. Meanwhile, the financial community continues to rely on the data, causing concern with experts like Kirthi Kalyanam, associate professor at Santa Clara University California and co-author of a landmark study on traffic ratings. "I think that [ratings] may be, unfortunately, highly misleading," he says.