Dadblog

Comscore have released an interesting study into how cookie deletions affect site metrics. Though it’s got some very good stuff in it, I’ve always been a little sceptical - obviously, as a panel-based firm, it’s in Comscore’s interests to emphasise the problems of cookie-based reporting.

But the conclusions are pretty, well, conclusive. Their conclusion is that “on average” there were 2.5 distinct cookies for each site in a particular month, suggesting that cookie-based stats are inflating user numbers by 2.5 times.

I’ve seen this figure before in other contexts, and there’s something I didn’t understand about it which now becomes clear. It’s the serial deleters who really inflate your numbers. The Comscore study thats that only 31% of users delete their cookies every month. That suggests that 69% of users don’t delete their cookies. Of the 31% of users who delete their cookies, there was an average of 4.5 cookies per site, suggesting these users delete their cookies roughly once a week, or 4.5 times a month. So each of these users is being counted as 4.5 uniques by cookie-based analytics.

But, in the Comscore survey, 7% of users had an average of 12.5 cookies, suggesting they deleted their cookies that many times in a month. So each of *these* users is being counted 12.5 times in a given month.

So, let’s say you have 100 real users. 61 of them don’t delete their cookies, so that numbers solid. Let’s say 32 of them delete on average twice a month, so they get counted as 64. And the final seven delete their cookies 12 times a month, so they get counted as 84. 61+64+84=209. Ouch. And if you’ve got a very techie audience, those cookie-deleters are going to be more prevalent. Double-ouch.

Back to the drawing board for some sites, I reckon.