• by Joe Mandese  @mp_joemandese, September 9, 2005

In a development that could influence how advertisers and agencies plan online media buys, a major provider of online audience research today will announce a breakthrough that enables it to produce more realistic audience reach and frequency estimates for online ad campaigns. The method, developed by Nielsen//NetRatings, utilizes a process known as data integration to reconcile the discrepancies that exist between the kind of panel-based sample research done by companies like NetRatings and comScore, and the so-called census-based audience estimates that come directly from a site's or portal's own server logs.

The approach, which already is being used by NetRatings in Italy, and has been submitted to the Media Rating Council for review as a potential service in the U.S., has found that there is a substantial overstatement in the reach and an understatement in the frequency delivered by many online advertising campaigns.

Advertisers, agencies and publishers who don't take this into account, says Manish Bhatia, senior vice president of product development and measurement science at Nielsen//NetRatings, are not reaching the numbers of users they think they are reaching, and they are paying considerably higher advertising costs for those they do reach. They're also reaching them more frequently than they may have known, or wanted.

The new method, which was celebrated by ad researchers at the Advertising Research Foundation's and ESOMAR's recent Worldwide Audience Measurement conference.

The major breakthrough, Bhatia told MDN is that NetRatings has been able to develop a system for factoring the number of times users delete cookies and therefore are identified as multiple users by publishers, portals and search engines as so-called "unique users."

According to NetRatings' research, that overstatement is significant. The company estimates that the overall market overestimates unique audience counts by about 27 percent. That overstatement can range from as high as 46 percent for search engines, portals and community sites to as little as 9 percent for travel sites.

While not yet fully vetted within the U.S. research community, such forms of data integration are becoming increasingly accepted by advertisers, planners and buyers as a solution to media research problems they cannot overcome in other ways.

Such disparities have long existed for magazines and newspapers, which derive their planning data - including readership and demographic data - from panel based surveys conducted by Mediamark Research Inc. and Simmons Market Research Bureau, respectively, which must be reconciled with the publication's audited circulation data.

On the Internet, companies like NetRatings and comScore provide the panel data, while site server logs provide the census data, which is frequently audited, but which according to NetRatings is greatly exaggerated in terms of unique users.

The development also has potential for influencing the development of the next generation of TV audience research. A number of players are pushing forward to develop so-called census data from the burgeoning digital set-top universe developed by cable and satellite TV operators, and devices like TiVo. While those set-tops produce empirical data on TV usage behavior - clickstreams that are analogous to Web server data - they are unable of providing individual TV user estimates the way Nielsen Media Research's panel-based methods do.

Nielsen has been trying to develop a set top-based system, as well as competitors including Taylor Nelson Sofres, and erinMedia, a company that has filed suit against Nielsen allegedly for blocking its entry into the market.

It's become clear to many observers that if a census-based database of TV viewing is to emerge it will have to be integrated with a panel-based approach as well, though erinMedia is also developing modeling techniques to get around that.