Wednesday, April 18, 2007

stats.datag.com How many visitors does your site really have? If your analytics rely on cookies, you may not know.

On Monday, the online audience measurement firm comScore Networks released new research that may make online publishers rethink some key assumptions about their online audience.

Almost a third of Internet users delete "first-party" cookies from their computers at least once a month. Also, seven percent of users delete these cookies four or more times per month, comScore found. (First-party cookies are those left by the site itself, not by advertisers and ad networks. ComScore also studied "third-party" cookies and found similar deletion patterns.)

These findings are enormously significant because server log analysis software and most other Web analytics systems rely on cookies to track unique visitors to a site.

ComScore's research focused on "one prominent Web property," which the company is not naming. Because of cookie deletion and other factors, the average user of the site received 2.5 cookies from that site in a month.

If this site tries to measure its usage via most Web analytics systems, there are two significant implications:

• The site's count of monthly visitors is 2.5 times as large as its real number of visitors.
• The site is underestimating the loyalty of its audience (via measures such as visits per month per user, page views per month per user or timespent per month per user).

ComScore's press release focuses mostly on the first implication: that sites are overestimating the number of unique visitors. But for publishers such as news Web sites, the issue of underestimating loyalty is just as important. If your site relies on server logs or cookie-based analytics systems to measure the number of users and their loyalty, you may have a completely misleading picture of your patterns of usage.

Assume that comScore's data extends to a site like yours (which may or may not be true). You might be thinking you have a healthy user base, but they aren't spending as much time on your site as you would like. So you might focus your staff's energy on trying to increase the loyalty of people who visit the site occasionally.

ComScore's study would suggest, instead, that you may have a relatively smaller number of users, but they're more loyal than you thought. In which case your biggest challenges are how to make your site relevant to non-users and how to attract their attention.

ComScore's findings are not entirely new. Back in 2004, market research firm Belden Associates concluded that cookie-based measurement was overstating sites' audience reach while underestimating the frequency of visits. (An article I wrote for NAA in 2005 cited other similar findings.) Still, comScore's data are a helpful reminder that cookie-based measurement tools -- including industry-leading systems such as those from Omniture and SageMetrics -- may not be counting visitors and measuring loyalty accurately.

The findings suggest that Web sites large enough to be monitored by comScore and Nielsen//NetRatings -- which estimate Web usage based on "panels" of online users -- may get more accurate measures of audience size and loyalty from these services than from their own analytics systems.

If your site is not large enough to be monitored regularly by comScore or Nielsen, and you're relying on server log analysis or other cookie-based analytics tools, you would be wise to assume you have fewer unique users and more loyalty than your current metrics indicate.

E-mail this item | Add/View Feedback (1) | QuickLink this item: A121682



E-Media Tidbits Archive

View items published between:		and
	(MM/DD/YYYY)		(MM/DD/YYYY)

MAIN | Back to Top