Quantcast this week will start offering people-based traffic counts for its sites, a hybrid of panel-based data and cookie-based measurement, using a statistical formula to account for the inconsistencies often cited when cookies are used to measure web traffic.
ComScore and Nielsen Online use panel data to figure out not only how many people are visiting websites but also who those people are. They argue panels are necessary because they provide a clear picture of demographics, which are essential for media planners. But critics say that as online media continue to fragment, panels simply cannot be considered reliable for measuring small and medium-size sites.
Publishers, meanwhile, tout audience numbers gleaned from their own servers, using analytics products from companies such as Google, Omniture or Coremetrics. These cookie-based tracking systems tend to report higher numbers than ComScore and Nielsen because if people delete their cookies, they're counted again as new users. If users access a site from multiple computers, they're counted multiple times. Alternatively, if multiple people use the same computer, they're counted only once.
But publishers hoping the new service will prove their numbers right are not likely to get their wish.
"There's no one answer whether our data will be higher or lower," said Adam Gerber, chief marketing officer at Quantcast. "Panel services base the audience on their sample. And there are many sites where the sample is accurate, but there are other sites where the panel may not be sufficient. ... It all depends on what biases are impacting the panel-services estimates and whether they're undercounting or overcounting traffic."
Quantcast is using statistical analysis to account for cookie deletion and the other issues typically associated with census measurement, extrapolating demographic data from its 1.5-million-person panel. It claims cookie deletion varies widely depending on the type of site. An e-mail site that a person visits daily or multiple times a day will have high variance between server and panel data. A reference site such as About, on the other hand, has fewer frequent visitors and is less likely to be affected by cookie deletion.
"On the face the claim sounds good ... but show me that you can actually get into the market space with a lot of established players," said Vipin Mayar, global exec VP-data and analytics at MRM Worldwide. He said the two parties that could be 800-pound gorillas in the space are Google and the internet-service providers, although neither has indicated it will go to market with a measurement tool (and both could face conflict-of-interest or privacy issues).
No matter what, audience measurement is seeing progress. A year ago, the Interactive Advertising Bureau called for better clarity, going so far as to advocate for census-based traffic counts. Since then, both ComScore and Nielsen are going through audits from the Media Ratings Council, and Quantcast recently announced it would allow the MRC to audit its panel.
Chris Parkin, senior director at Omniture, said a few smart publishers are already merging panel-based or CRM data with data from their own analytics tools. Quantcast's multipronged approach to audience measurement is the way to go, he said, but "there are pretty significant barriers to entry."
The new data is available to publishers that participate in Quantcast's Quantified program, meaning they have placed a small bit of code that tracks visits on their sites. "We recognized from the start that panels were not the end of the story," said CEO Konrad Feldman. "They're useful, but you really need a direct-measurement solution as well."