Eric Peterson made a comment in the Yahoo! Web Analytics Forum about the recent ComScore announcement that Uniques are overcounted an average of 2.5 times using cookie based counting. I'll give my take after briefly going over the reaction some are having about the major announcement from ComScore:
Eric Peterson wrote:
".. does the knowledge that the estimated reach for a particular property could be off by a factor of 2.0 or 2.5 bother you at all? Or is it okay to know that the real reach for a site is somewhere between what the property tells you and what the panel services tell you?"
Judah Phillips replied:
"..A well-timed business press release serves a purpose that is not necessarily scientific. In this case it all may be related to ComScore's IPO."
Does ComScore's announcement help the company?
Here's the ComScore results for average 1st party cookie deletion:
comScore Cookie Deletion Analysis – 1st Party Cookies* December 2006 Source: comScore, Inc. | |||
Number of Cookie Deletions/Resets | Percent of Computers | Avg. No. of Cookies per Computer | Percent of Cookies |
Total Sample | 100% | 2.5 | 100% |
1 or more | 31% | 4.7 | 58% |
4 or more | 7% | 12.5 | 35% |
On average your getting 2.5 times less Uniques than you think your getting based on this announcement; but if your audience is made up largely of smaller percentage of visitors who delete their cookies 4 or more times a month your audience may 1/12th or 1/13th the size you believe it is. Ouch! Double Ouch!!
And it's even worse if your basing Uniques on 3rd Party cookies and your audience is made up largely of people erasing their cookies more than 4 times a month; you real number of Uniques could be 1/14th of what you thought it was.
Doing a Search on Internet Retailer Unique Visitors shows what some large sites think their Uniques are vs. what they may really be:
Best Buy 17.3 million shoppers (uniques)in December 06 - with the corrected "Comscore" number that would really be 6.92 million shoppers!
Circuit City 10.2 million shoppers (uniques) in December 06 turns out to really be 4.1 million shoppers.
SonyStyle 4.9 million "shoppers" ends up being 1.96 million shoppers.
What do I think? The current way of overcounting Uniques benefits big retailers by inflating their visitor count. The corrected numbers are much more believable and feel right - but I don't think anyone who sells something on the web and puts a value on the number of visitors they're getting will be in any hurry to divide their web analytics sanctified Uniques by 2.5.
Comment Preview