Web Firms May Vastly Inflate Claims of ‘Hits’
The fortunes of many Internet companies depend on how many people look at their World Wide Web pages and online advertisements.
But with no guidelines or regulations about how Internet traffic is measured, many Web sites frequently inflate by 50% to 100% the traffic figures they give to advertisers, investors and the public, executives say.
“The industry has turned a blind eye on this whole issue of where the data come from,” said Tim Meadows, senior vice president at Nielsen NetRatings, one of the leading Internet market research firms.
The measurements are most critical for content sites, which are depending on online advertising to climb from $2.8 billion last year to $22 billion in 2004, as Forrester Research projects it will.
“When the [traffic] numbers start to go down, there will be a lot of temptation to cheat,” said lawyer Boris Feldman of the Silicon Valley firm Wilson, Sonsini, Goodrich & Rosati, who advises Internet companies.
“It’s a side of the industry that no one wants to talk about,” said Darren Ting, vice president of Net Power in Las Vegas, which owns the Banner Brokers advertising network, which sells ads on 1,500 Web sites.
With more than 200 million active Web sites, getting a handle on true customer traffic is an overwhelming task.
Two leading Web traffic survey firms are Nielsen NetRatings and Media Metrix. Both firms install tracking software in personal computers of consumers to gather raw data in a fashion similar to Nielsen’s television ratings.
Although they routinely publish data about the 50 most popular sites by category, rankings of less popular Web sites are rarely seen. The research firms charge for most of their information, and with only a few hundred companies as clients, audience figures for less popular sites aren’t publicly estimated.
As a result, the sites beyond the 2,000 or so most frequented destinations simply tell advertisers or investors what their visitor traffic is based on what their in-house computers automatically record. And to a large extent, what those logs reveal is a matter of interpretation.
Many programs that analyze the logs don’t weed out traffic hits that are generated within the company or by robots that are trolling the Web on behalf of Internet search engines, let alone prevent deliberate fraud.
“Some heavily trafficked consumer sites have significant amounts of robot traffic, as high as 40% of the total,” said Patrick Monahan, vice president at BPA International, a New York auditing firm. And Vice President Scott Hanson of auditor ABC Interactive said he has seen a 75% robot hit rate on some Web sites.
ABC Interactive is an affiliate of the Audit Bureau of Circulation, the venerable newspaper and magazine circulation tracker. But unlike the situation with more-established media, no standard for what should count as traffic has emerged, and that makes reasoned decision-making difficult for commercial partners, advertisers and dot-com stockholders.
“There is an FASB [Financial Accounting Standards Board]--but there is no WASB, a Web equivalent,” said Bob Page, chief technology officer of Accrue Software Inc., which sells programs that help companies analyze the habits of the surfers who visit their pages. “What is counted depends largely on the site.”
Several hundred second-tier Web sites do pay for outside audits. Even then, auditors said, their clients frequently argue over how the traffic should be counted. One sore point has to do with Internet search engines that might briefly touch a Web site while rapidly screening to list sites for a keyword search.
The operator of that Web site will frequently demand that each search engine “hit” be counted as a customer visiting his site. Another typical dispute centers on whether a user requesting a Web page that has 10 graphical elements on it can be counted as registering 10 hits, as some companies want, or just one.
“Those people are short-term greedy,” said Stephen Collins, chief financial officer of DoubleClick in New York, the largest advertising network on the Web. DoubleClick sells ads for Microsoft, AT&T; and others and posts them on Web sites ranging from AltaVista to U.S. News.
Not that an absence of credible Web traffic numbers stops companies from bandying about the ones that sound best.
Consider TheGlobe.com, based in New York, which offers tools to let users construct their own Web pages and hosts other services such as chat forums. In May, the company raised $70 million in a secondary stock offering, and three of its executives sold $40 million worth of shares. TheGlobe.com put on the cover of its stock prospectus that it had “10.2 million users.” It attributed the traffic figure to ad agency DoubleClick.
But that figure was wrong. TheGlobe.com tersely retracted it a month later, after the stock sale and to little public notice. In fact, TheGlobe.com was drawing only about 3.2 million visitors, according to Media Metrix, making it an unspectacular No. 44 Web site for the month of April 1999.
DoubleClick says it miscalculated the traffic because of human error and it faults TheGlobe.com for putting it in a securities document without permission.
Media Metrix said another Web site showed a sudden surge in user traffic--but no increase in the total number of pages the users were looking at. When the company investigated, it learned that a corporate directive at the Web site instructed its staff to increase the number of recorded users, one way or another, said Media Metrix Senior Vice President Andy Fessel.
“We can easily have [traffic] variances that are [inflated by] 100%. It can range higher than that,” Fessel said. “The theme is ‘increase.’ There are some tremendous pressures” to satisfy public investors and venture capitalists.
“There are going to be accounting restatements” because of inflated Web traffic figures, attorney Feldman said. “There will be hundreds of shareholder lawsuits.”
The major Web traffic survey firms use techniques similar to the Nielsen TV ratings. Media Metrix, Nielsen and PC Data mechanically track from 50,000 to 100,000 Web surfers at home and at work to estimate U.S. World Wide Web traffic.
But their surveys do not include where 30 million America Online surfers are going inside AOL’s proprietary sites. And estimates by the three often disagree with each other, probably because of differences in how their panelists are recruited.
Shaky as Web audience figures are, though, the reliability of advertising viewership and “click-through” rates is probably worse.
The major Web market research firms generally don’t track the number of viewers who actually click on a banner ad to learn more about the advertiser. Instead, ad buyers depend on the reports of Web sites or ad networks that distribute ads to multiple sites. But frequently those ad figures aren’t audited, resulting in widespread exaggeration, experts say.
Exactly how inflated those figures are becomes clear when an auditor or audience measurement firm drops by to compare notes with a client.
“We will have people that are unhappy with the numbers,” said Monahan of BPA International, a nonprofit group founded by publishers in 1931 that began doing Web site traffic audits in 1996. BPA scrutinizes computerized logs of Web sites to verify advertising “click-throughs.” Because the auditors are hired hands, their clients aren’t forced to publicly reveal the unhappy results.
“We have had companies not release [our] audit figures because the numbers are lower than what they’ve said publicly,” Monahan said. “We don’t know what they’ve told advertisers. [But] as they turn white, you get a pretty good idea it’s a wide swing.”
Fraud Is Easy for Hackers
One problem has to do with payments made to Web sites that steer customers to online ads and get paid for each click-through. It’s easy for hackers to set up their own Web site, apply to an advertising network and start taking in banner ads. Then the hacker can run programs to send false clicks through those ads, over and over, getting paid a few cents for each hit.
Most ad networks say they check for the most obvious fraud: repeated clicks coming from the same Internet address every 10 seconds, say, or ads that every single viewer appears to be clicking on, for a 100% click-through rate. But the hackers are getting more sophisticated, and there aren’t enough security professionals scrutinizing Web logs by hand.
Consider Jason Pearsall, 19, a University of Massachusetts student. Pearsall and some friends worked up a fake click-through program. The program used an Internet chat room with about 100 people in it. Pearsall said it was simple to fool Web sites into thinking that everyone in the room was referring everyone else in the room, so everyone got multiple bounties automatically.
Pearsall said that he and his friends made about $1,000 each by using the program, and that he believes similar scams are widespread. Pearsall has since been hired as a consultant to work with Banner Brokers, among others.
“As long as you have anonymous surfing, the technical means to cheat will almost always remain a step or two ahead,” said Pleasanton Internet ad consultant Mark J. Welch, who said he found 50% nonhuman traffic on several pay-per-click networks.
To overcome this unreliable reputation, the nonprofit Internet Advertising Bureau has tried, and failed, to set online ad performance measurement standards since 1997.
The issue is seldom discussed in public, and companies are very sensitive about it. At the forefront are the ad networks like Carpinteria-based ValueClick, which charge advertisers based on the number of click-throughs. ValueClick President Earle Malm refused to estimate how many of the clicks it charges for are not human, saying only that the figure is less than 50% and that it is lower today than in the past.
Some Firms Demand Greater Accountability
The shift away from basing ad rates on the number of viewers who see an ad and toward payment per click-through has been driven by advertisers, including consumer products giant Procter & Gamble, who are demanding greater accountability for their spending.
But with increasing evidence of fraud, advertisers are pushing for even more restrictive plans, including “pay-per-lead,” in which referring Web site gets a payment only for each viewer that both clicks and indicates more interest, say by leaving an e-mail address to be notified of future sales. Other Web sites are even resorting to “pay-per-sale,” in which the referring site gets a cut only when the viewer is billed for an order.
“What really matters to an advertiser is not so much the volume as the quality of the users,” said Charlene Li, a senior analyst at Forrester. All told, pay-per-click and other pay-per-performance models should increase from 15% of Web ad sales last year to 50% by 2003, Li said.
But paying only for performance doesn’t fix the whole problem.
ValueClick, which is the largest pay-per-click ad network, said in its initial public offering filing last month that it rejects 80% of the Web sites that apply for membership because of “inappropriate content, insufficient traffic, illegal activity and fraudulent clicking activity.”
“From time to time we have had to resolve differences between our measurements” and counts of human traffic by clients, ValueClick said in its filing. “Any significant dispute over the proper measurement of clicks or other user responses to advertisements could cause us to lose customers.”
The problem is compounded because perpetrators face almost no punishment. Networks like ValueClick say they kick fraudulent sites off their system and deny payment for bogus clicks, but that’s not much of a deterrent. Because each cheater takes very little money, on average, prosecutions and lawsuits are unheard of.
Microsoft this year ended its own pay-per-click ad network after “some problems with fraud” and other issues, said product coordinator Justin Slane.
Even modest efforts to compile databases of known hustlers have gone nowhere because ad companies don’t want to share hard-won information with their competitors. They also don’t want to shake public confidence in the entire Net advertising system.
“I suggested about a year ago a public forum for reporting cheaters,” John Ferber, co-founder of the Advertising.com network, wrote in a chat-room post last spring. “After discussing it with more people in the industry, we all decided that it’s much better to keep your dirty laundry hanging inside than outside.”
