The Impact of Cookie Deletion
December 4, 2007 – 12:26 am
I’ve come across a comScore study released in June of 2007 which attempts to measure the impact of cookie deletion on what they’re calling “site-server” and “ad-server” metrics. I’ll let everyone download / read the whitepaper at their own leisure but I’ll summarize their conclusions and implications:
Conclusions
- 31% of US users delete their first-party cookies in a month
- First-party and third-party cookie deletion rates are generally similar
- Computers with active security programs (i.e. Mcafee or Norton) experience somewhat higher cookie deletion rates, but nothing incredibly significant
- “Serial cookie deleters” (users who often delete cookies after every browsing session) have a profound impact on inflating analytics tools that use first / third-party cookies for tracking (account for 7% of computers, but represent 35% of observed cookies)
- Analytics tools using cookie-based tracking measures can overstate true number of unique visitors by a factor of up to 2.5
Implications
Cookie deletion leads to following inaccuracies in site-centric measurement:
- Overstatement of unique visitor counts
- Understatement of repeat visitor counts
- Understatement of conversion rates
Mike’s Take…
Alright, we’re all aware (or should be by now) that cookie deletion represents a problem for the Google Analytics’, Gatineau’s, Coremetrics, Omniture’s and WebTrends SDCs of the world, but studies like comScore’s don’t mean we all should start running for the hills just yet.
The number one problem with studies such as these is it truly takes people’s eyes of the ball of what they really need to be paying attention to trends! What the comScore study notes (which is important) is that there’s a level of inaccuracy about measurement tools that use cookie-based tracking. But I would argue that web analytics isn’t about accuracy, it’s about reliability! Say I know for a fact that 31% of my audience on average possibly deletes their cookies per month every month. Then what I have is a predictable margin of error. If I know this is the behaviour my audience exhibits then I can just note that in an analytics report and be on my way. Thus, web analytics is still reliable.
And folks here’s the plain and simple truth: when it comes to the Internet, there really isn’t one best way of tracking your visitors.
comScore uses a panel to capture its measurements. You sign up to be a part of the panel, often because of an incentive like a contest or in exchange for antivirus software. In exchange for whatever the incentive is, comScore is allowed to place What are the big problems already that we see here?
- By definition, with comScore you are only sampling an audience. Samples are great and there are plenty of statisticians who’ll argue to tell you how valid they can be using some pretty advanced mathematics. But as they would also tell you, sites require statistically significant audience sizes in order for a random sample of the audience to accurately represent the whole - this unfortunately is not the case for many small guys. It’s been said that in Canada, don’t even bother with comScore unless your audience size is above 100,000 monthly uniques (I personally think you’d need around 500,000).
- Getting people to do anything with narrow incentives automatically creates a bias in a sample. There’s a certain type of person who goes for these incentive based activities, I personally would never be one of them and I’m sure I’m not alone. Therefore, you’re only getting a certain segment of your online audience when they visit - that’s no good!
- Downloading anything onto a computer is traditionally a personal computer thing only - not something you do at work. Therefore if you’re a site that relies on a business audience (Toronto Stock Exchange for example) then you might be stuck. Yet another bias!
And yet, many people rely on comScore (especially advertisers!) and I can’t say that I blame them, after all, what else is there (in Canada) for competitive analysis?
The point is, comScore itself has some very large troubles too but I won’t bash those guys either. When it comes down to it I always tell people one thing, in what other marketing channel could you ever get the depth of information that you are able to get when using a web analytics tool? It just doesn’t exist…that’s partially why so many marketers are looking to spend more online, you can finally directly prove the worth of marketing efforts with these tools.
3 Responses to “The Impact of Cookie Deletion”
“Analytics tools using cookie-based tracking measures can overstate true number of unique visitors by a factor of up to 2.5 ”
Per day , per Month , per year , per what ?
Different audiences have different habits.
Comparative results may have some meaning.
From MY strict IT point of view , web analytics “absolute unique visitors” per month is an Imaginary number.
Thanks for your time
By Tsichles on Feb 20, 2008
In response to Tsichles post…
Thanks for posting!
I believe the study was referring to a monthly timeframe but you can always download it to be sure.
Absolute unique visitors you’re right is not a great metric to rely on if what you’re looking for is ACCURACY. However, if what you’re looking for is RELIABILITY then there’s theoretically nothing wrong with using it.
The difference between the two obviously being that accuracy is important if we intended to use analytics for accounting - which we don’t!
What we’re usually using web analytics for is optimization which always means we’re looking at comparative results hence the metric has meaning despite being somewhat “imaginary” as you suggest :).
By mike.sukmanowsky on Feb 20, 2008