• by Gavin O'Malley  @mp_gavin, April 5, 2005

A recent Jupiter Research study concluding that four out of 10 Internet users delete cookies at least once a month has been met by the Web analytics industry with everything from skepticism to outright disbelief.

On Monday, Eric Peterson, the Jupiter analyst who authored the controversial study, responded via blog, offering doubters a combination of survey and direct panel measurement data from Nielsen//NetRatings that support his findings. According to Peterson, two Nielsen executives--Michael Saxon, vice president of products, and Richard Goosey, vice president of research and development--provided him with the data sometime last week.

Critically, one Nielsen survey found that 43.7 percent of 9,492 respondents surveyed said that they had deleted cookies in the last 30 days. "We found that only 39 percent of people were deleting cookies," Peterson commented in a phone interview.

Nielsen's Michael Saxon Monday confirmed the accuracy of the data that Peterson cites on his blog, adding that Nielsen first presented the information last year. "I think it's a very good summary of our findings," said Saxon.

Peterson's study--the first Jupiter study to examine how Internet users react to the cookies that wind up on their personal computers--rebuffed a generally accepted belief that Web surfers leave cookies alone. "It was commonly assumed, before this study, that users didn't have the sense or the inclination to fool with cookies," Peterson said upon the initial release, "so advertisers and marketers didn't factor the possibility into their tracking and targeting measurements."

When Nielsen compared visitor retention rates between its Web analytic application--SiteCensus, formerly known as RedSherriff--and their NetView measurement panel, Nielsen determined that the cookie-based retention metric reported half the number of returning visitors as the measurement panel on five-out-of-six sites.

What's more, when Nielsen leveraged their measurement panel--a direct measurement from participant desktops, able to report back when cookies are being set--to examine seven different portal and publishing Web sites, Nielsen reported one-month unique cookie deletion rates that ranged from 7 to nearly 50 percent for both at-home and at-work audiences.

Peterson qualified Nielsen's findings by noting that the company has a "vested interest in this data coming to light (they recommend an integrated solution using census and panel-based data for Web analytics to resolve these issues of reach and frequency.)"

All the same, when Nielsen looked at Google.com in July 2004, it found that nearly 25 percent of its measurement panel received a new google.com cookie during the study, and nearly 15 percent of that group received multiple cookies, resulting in a 55 percent overestimate of unique cookie counts, according to Peterson.

Peterson wrote in response to the Google study and what it could mean for analytics accuracy: "Of all of the members of [Nielsen//NetRatings]'s panel ... only one quarter are deleting and blocking their cookies. And this activity is only resulting in a 55 percent inflation of unique cookie counts, which, depending on which analytics application you use, would potentially increase your monthly unique visitor count by 55 percent while potentially reducing the accuracy in your retention metrics by 25 percent."