Welcome to WebmasterWorld Guest from

Forum Moderators: DixonJones & mademetop

Message Too Old, No Replies

New Study - Cookie tracking flaw

3 in 10 US delete their cookies

10:38 am on Apr 17, 2007 (gmt 0)

Senior Member

joined:Mar 8, 2002
votes: 0

Comscore has a new study out, yesterday

Frequent Cookie Deletion by 3 out of 10 U.S. Internet Users Leads to Overstatements in Audience Sizes by a Factor as High as 2.5 [comscore.com]

A claim I find hard to believe. The report makes for immensely interesting reading, but I challenge the baisi of the research. The research used two "sample groups" - a big site and am "ad network". It is this second sample that I would hazard is causing the problems - Ad-networks shouldn't be dropping cookies should they!? So maybe spyware software has a very real chance of blocking this third party's cookies.

Doesn't mean 3 out of 10 US users delete first party cookies from REAL sites does it?

There are indeed several problems with cookie only based tracking - something we've known four four years [webmasterworld.com] but other methods have other issues.

A good tracking system, though, needs to provide an iterative method that combines several tracking mechanisms I guess. But if a user really does delete their cookies regularly 30% of the time, then this pulls into question the primiary iteration - but only for the "all time unique" metric. I guess the only way to truly verify a unique returning user is a handshake - a formal login. that's something that webmasterworld does quite well. It looks for the cookie. If the cookie is still there, then all is well. If not, then after a while, you need to log in to continue.

But I am not so sure that tactic is being used for tracking as much as bot-blocking.

Anyway - interesting report by Comscore [comscore.com] - but does anyone know the third party ad network that they used in the story?

12:52 pm on Apr 24, 2007 (gmt 0)

New User

10+ Year Member

joined:Apr 24, 2007
posts: 2
votes: 0

It seems high. But it is a press release. It's meant to increase awareness of the brand and drum up business. As a study it is extremly suspect, and as presented it is weak at best.

I haven't been able to find any more data on the study than what is presented in the release. If anyone can find more info on the study, please post it.

2:34 pm on Apr 24, 2007 (gmt 0)

Senior Member

joined:Mar 8, 2002
votes: 0

Looks like the IAB doesbn't agree with Comscore and Nielsen's ways of tracking either from what I am gleaning [adweek.com].

[edited by: Receptional at 2:36 pm (utc) on April 24, 2007]

12:10 pm on May 3, 2007 (gmt 0)

New User

10+ Year Member

joined:Dec 17, 2006
votes: 0

The comScore report is indeed dubious, and there is a great deal of debate going on between supports of "panel-based" traffic metrics (people who seem anxious for the comScore IPO), and people who know what they're talking about. :P

Their reports and rhetoric purposefully confuse the line between cookie-based analytics, 3rd party JavaScript analytics, and server logs in an effort to keep online media buyers confused about something that's not so confusing.

Below are the details that should provide plenty of fuel for discussion... but, for those not inclined to spend time with the details, here's the short story...
- Only 8.9% of the weekly new users either deleted or refused cookies
- Google Analytics was only 1.5% high in their report of absolute uniques
- As I suspected, the number of users who manage cookies is dwindling

Unique User Test

Here are the preliminary results of a hard-nosed unique user and cookie-deletion test that has run for one week. The cookie results are surprising, given that visitors to our site ( a popular conspiracy theory discussion board ) are those most likely to either refuse cookies, or delete their cookies daily.


We used the following criteria, in this order, to determine that we had a new, unique user.
1) New visitor is not a site member
.......(logged in members are a known entity)
2) New visitor with no previous cookies from our domain
.......(either never visited, or deleted cookies)
3) The visitor came in via some page other than the home page
.......(regular users tend to bookmark the home page)
4) The visitor came in with a referral from another domain
.......(regular users tend not enter the site from an off-site link)
5) The visitor had a "User Agent" identifier
.......(some bots now have blank user agents)
6) The visitor's User Agent was not a known spidering bot
.......(we compared against a list of 50 known bots)
7) The visitor views pages at "human speed"
.......(rapid-fire visits are likely unknown bots)
8) The visitor's combination of IP, OS, and User Agent is unique
9) If the user logs-in (a member), they are not counted

If the above was met, we logged the user and wrote a new cookie with a one year expiration date.

So far, in 168 hours (one week), we tracked the following statistics:

Unique users recorded ---> 121,718
Combined users deleting & refusing cookies ---> 10,815 (8.9%)
Number of repeat users deleting cookies ---> 7,439 (6.1%)
Number of repeat users refusing cookies ---> 1,186 (0.9%)
Number of one-shot visitors refusing cookies ---> 2,190 (1.8%)
Number of members visiting for the week ---> 3,081 (124,799 uniques)

Google Analytics "Absolute Unique First Time Visitors:" ---> 123,498 (off by 1.5%)
.......(for the same time period as the user/cookie test)

We're going to continue running this, and compare 30, 60, and 90 day data... I suspect the gap between our numbers and Google Analytics will widen and see a slight increase in cookie deletion.

1:48 pm on May 3, 2007 (gmt 0)

New User

10+ Year Member

joined:Apr 24, 2007
votes: 0


Nice work. I'm interested to see the long term results.

What are your total site traffic numbers?

8:54 pm on May 3, 2007 (gmt 0)

New User

10+ Year Member

joined:Dec 17, 2006
votes: 0

Last month we had over 5 million page-views, on about 1.4 million visits.