Online ad squabble nears resolution

A standards body will issue new rules for counting online ad "impressions," possibly affecting hundreds of millions of dollars in annual payments from advertisers to Web publishers.

Stefanie Olsen Staff writer, CNET News
Stefanie Olsen covers technology and science.
Stefanie Olsen
8 min read
A key Internet standards body is scheduled next month to issue new guidelines for counting online ad "impressions," a move that could affect hundreds of millions of dollars in contested annual payments from advertisers to Web publishers.

At stake are the billing methods and ultimately the credibility of the tarnished online advertising market, where inconsistent data for years has led to conflicts between ad buyers and publishers.

According to industry veterans, about 10 percent of all online advertising charges wind up in dispute--about $600 million worth in 2001, based on projected online ad sales of $6 billion to $8 billion.

In the past, publishers frequently received the benefit of the doubt in such cases, according to ad executives. But the advertising market has weakened substantially, giving advertisers the upper hand in such disputes, experts said.

"Publishers' and advertisers' economic incentives are misaligned," said Ebrahim Keshavarz, former vice president of product development at DoubleClick and general partner of EbTide Consulting. "Advertisers want to see a minimum number of ad impressions show up, and publishers want to see the most...Over time the (standard) is most likely going to move to the advertiser's view of the world because that's where the money and power is."

The Interactive Advertising Bureau (IAB) next month plans to issue a proposal to reconcile ad-counting discrepancies.

According to the IAB, new standards could go a long way to improve industry relations and cast off negative beliefs about the embattled sector. After years of unbridled growth, Web advertising has slumped because of the economic downturn and doubts about its effectiveness among some advertisers. At the heart of many concerns are the mass amounts of data the Net is capable of churning out and the countless ways those numbers can be interpreted.

This attitude sharply contrasts the Net's heyday, when the medium was hailed as a revolutionary one-to-one marketing channel, capable of extremely detailed reporting on how consumers respond to advertising.

In the afterglow, the reporting became an albatross. Metrics such as "click throughs," or the number of times someone clicks on an ad, steadily plummeted, and advertisers began scrutinizing the value of investing online. As a result, industry leaders started resizing ads and running tests on the medium for its branding capabilities.

Now industry executives are aiming to root out strife that runs deep within the fundamentals of the ad reporting. Chief among the troubles in counting ad impressions is the different ad-serving systems used by publishers and agencies.

Various systems count impressions at different points in the ad's path from the server to the site, causing discrepancies in reporting. Companies also have various means of factoring in when so-called Web crawlers request or servers cache a Web page. In fact, ad executives say the only thing that's consistent in reporting is the inconsistency.

Companies such as major packaged-goods advertisers "are reluctant to come onboard because they say, 'Until you guys agree on the numbers, this is a nightmare medium,'" said John Keck, media director of the interactive division of Foote Cone & Belding, a San Francisco-based advertising agency.

Which clicks count?
In its attempt to reconcile the differences, the IAB has hired PricewaterhouseCoopers, which is collecting information from the technical and sales staff at 11 major Internet companies, including third-party ad servers, portals and "destination" sites.

The consultants are examining the processes for measuring ad impressions, page impressions, clicks, unique users and total visits. By looking at the most common practices, the industry group hopes to set standards called ad campaign measurement guidelines by mid-November. Robin Webster, chief executive of the IAB, would not disclose the names of participating members.

Gartner analyst Denise Garcia says that in a fair and just online world, advertisers shouldn't have to pay for ads that viewers haven't seen.

see commentary

The proposal is still at an early stage and may change before its release. Nevertheless, a preliminary draft seen by CNET News.com signals that publishers may be forced to pay a steep price.

Specifically, the draft recommends recording an impression "as late as possible in the process of delivery...to the user's browser." That means millions of ads that are interrupted before appearing on a browser window, for whatever reason, would not count. The draft also recommends filtering out the effects of automated searches that inflate the total number of pages viewed.

In a step toward reconciling this difference, the IAB and Web metrics auditor ABCi on Monday released an industry-standard list of automated searches, known as spiders and robots, meant for Web sites to use as a de facto list of what to filter from ad and Web page reporting. The IAB also issued a glossary of Web advertising terms last week to set a common language for advertisers and publishers.

Publishers may come out ahead in at least one hotly disputed area: The draft hints at new rules for dealing with data-hosting techniques known as caching that have historically undercounted the total number of ads served.

But overall, the situation marks a reversal of fortune for Web publishers, which once enjoyed significantly more leverage with advertisers. A few short years ago, when inventory on sites such as Yahoo was sold out and ad dollars flowed freely, the numbers would regularly skew toward the publisher, ad executives say. In fact, contracts typically include legal language stating that the publishers' numbers--not the advertisers--count when figuring impressions.

But when the ad market dried up, advertisers suddenly had more power to dispute the publishers' numbers with their own data. Faced with losing an advertiser such as Procter & Gamble over a discrepancy of 1,000 ad impressions in today's grim market, for example, most publishers would consider deferring to the client.

"On the back end, advertisers want to know, did my ad get delivered?...(But) whose numbers do you use--third-party ad servers or publishers? There are multiple ways to measure an ad impression, and one is not necessarily better than another," Webster said. Standards "will help reduce the reconciliation costs between the agency and publishers and make the advertiser comfortable knowing that the numbers are good."

Currently, publishers often count an ad impression, or the delivery of an ad, when a line of code requesting the advertisement from a third-party ad server is sent with the delivered Web page. Conversely, advertisers, or their third-party ad networks such as DoubleClick's Dart, often count the impression only when the image itself is delivered.

The image's delivery can be hampered by any number of complications, including Internet congestion, the absence of a plug-in necessary to view the image, or the failure of the visitor to click off the Web page before the ad was displayed. As a result, the two numbers reporting ad impressions often don't match up.

"Without a standard, how are we going to collect money? Are we going to say that your third-party server is much better than ours, and sure, pay us $15 on that $2,000 deal? There's an enormous amount of room for error right now," said Sandra Baez, director of advertising operations at CBS MarketWatch.

Ad executives at MarketWatch say the company regularly defaults to the advertiser's numbers if the difference is less than 10 percent. If it's more, the company tries to negotiate a solution with the client.

Adding up the numbers
Most publishers say that the ad should be counted when the Web page is loaded, with or without the ad. Michael Zimbalist, executive director of the Online Publishers Association, compares this view to traditional media delivering a newspaper or magazine with ads.

"The publisher, by analogy, is responsible for putting the page up. When the ad is coming from someplace outside of the publisher's control, the publisher shouldn't be penalized for not delivering its content and audience as promised," he said. "It would be like an advertiser in a magazine saying, 'Yes, it is true that you have 100,000 circulation, but only 25,000 people turned to my ad.' If the delivery of the ad is out of their control, they shouldn't be penalized."

But agencies are quick to defend their advertisers' turf.

"Sites would rather count everything. But if we don't know if people saw (the ad), we don't want to pay for it," said Scott Sanborn, vice president of client services for iTraffic, an interactive marketing subsidiary of Agency.com.

To this end, advertisers or ad networks try to count the impression when the image is requested from the ad server. The logic is that if a person is surfing the Web using a slow, 28.8kbps connection, it may take several seconds for a page to load; in that time, the person could click off the page, refresh the site, or hit the back button before seeing an ad.

Discrepancies do not crop up as long as publishers serve their own ads and produce only one set of numbers. But advertisers and agencies are inclined to use a third-party ad server for two reasons: They can audit the performance of Web sites independently, and they can track a campaign across several sites at once.

"We would prefer not to have to rely on the site's numbers because a central server's system is consistent across all sites," Sanborn said. "I'll get the site's reporting and the central serving numbers, and the numbers don't match. Whose numbers do you trust, and what's the real story here?"

Another cause for squabbles lies in how ad servers and sites handle "caching," or the process of loading Web pages on a PC or server to save bandwidth costs.

Instead of loading a Web page each time it's requested, many company servers capture a Web page the first time it's downloaded. This allows many people in a company to view the page cost-effectively. But this makes it difficult for a publisher to track how often the ad is viewed when it could be delivered several times from the server but the Web page is only counted once.

In addition, companies have different ways of filtering and counting Web pages and ads requested by Web crawlers, or robots used by search engines such as Google to catalog content on the Web. Advertisers want to ensure that publishers and ad networks filter out the bots.

Although millions of dollars are riding on the resolution of these issues, some say they believe the biggest problem has been the long delay in establishing commonly accepted principles.

"We don't care what the standard is," Sanborn said. "I would rather not argue the specifics of is it cached or not cached or does it filter robots or not. I'd rather have an imperfect standard than none."

Others take a long-term view of new standards.

They're "important for the stabilization of the industry," said Benjamin Grant, vice president of technology at online magazine Salon.com. "It certainly hampers the development of the online ad industry when the folks that are doing the media buying don't feel confident in the numbers."