• by Wendy Davis  @wendyndavis, September 9, 2005

Nielsen//NetRatings is expected to unveil today a new way of counting the number of visitors at Web sites that takes into account the proportion of users who delete cookies.

The new method--currently available in Italy and Australia, and being tested in the United States--relies on combining information about cookie deletions from Nielsen//NetRatings' consumer research panel, with Web sites' raw visitor numbers, as determined by SiteCensus, Nielsen//NetRatings' analytics tool. Site Census collects information based on cookies.

Currently, there are two major sources of discrepancies between numbers from Nielsen//NetRatings' panel and its analytics tool, said Manish Bhatia, senior vice president of product development and measurement science at Nielsen//NetRatings.

The first is that Nielsen//NetRatings' panel excludes public computers--such as those in libraries or schools--leading to undercounting of visitors who access Web sites from such computers.

Another major factor is that the analytics tool relies on cookies to keep track of visitors. But this cookie-based methodology results in an overcounting of visitors who delete their cookies, because sites end up placing a new cookie on the same computer each time its browser lands on the site.

Now, Nielsen//NetRatings says it can combine data from its panel and its analytics tool to compute the deletion rates, and then combine them with the site's own numbers to generate a more precise estimate of visitors.

First, Nielsen//NetRatings determines whether it has placed a site's cookie on the same computer more than once during that month. Then, it can figure out the proportion of users who have deleted that site's cookies once during the month, twice in the month, and so on.

From that data, the company arrives at a calculation showing the ratio between users and cookies. Then it applies that percentage to the number of visitors the site's analytics tool reported, ultimately arriving at a more accurate count.

For example, if Nielsen//NetRatings' panel data shows that 40 users generate 60 cookies in a month for a particular site, that site's ratio would be 1.5 cookies per user. Therefore, if the analytics tool reports 100 cookies, Nielsen//NetRatings would divide that number by 1.5 to arrive at the total of about 67 visitors.

Bhatia said that Nielsen//NetRatings currently is able to calculate cookie deletion rates only for sites that use its analytics tool, SiteCensus, because Nielsen//NetRatings recognizes those cookies. While it's theoretically possible for Nielsen//NetRatings to identify new cookies placed by other analytics companies--Coremetrics or WebTrends, for example--doing so would depend on those companies sharing information with Nielsen//NetRatings, and would be a cumbersome procedure, Bhatia said.

Until this year, many in the online advertising industry assumed that few, if any, consumers were deleting cookies. But research released in March by JupiterResearch Senior Analyst Eric Peterson revealed that four out of 10 Web users delete cookies from their primary computers at least once a month. Other studies in the last six months have likewise shown that consumers are deleting cookies at rates few would have imagined even a year ago.

Cookie deletions would be expected to most affect analytics at sites that are visited most frequently. For example, if someone visits a search engine every day, but deletes all cookies once a week, that search engine might consider that same visitor to be four unique users, based solely on cookie data. But if that same visitor only goes to a travel site once during the month, it won't matter that he's deleted his cookies--he'll still just be counted once based on the one visit.

In fact, a Nielsen//NetRatings analysis of cookie deletion rates in Italy in July bears out this theory. Sites such as search engines and portals are the most likely to overcount their visitors, while sites visited less frequently, such as insurance sites, are least likely to overcount.