Commentary
The Impact Of Cookie Deletion On Marketing Attribution
- by Anto Chittilappilly , Op-Ed Contributor, April 11, 2012
One of the questions frequently asked about attribution is how much cookie deletion by Internet users affects its accuracy. This is an extremely valid question, because proper attribution of digital channels such as search and online display is very much dependent on these little pieces of code placed on users’ computers to collect the data associated with the touchpoints created by your campaigns. So without cookies, the attribution process that’s dependent on that touchpoint data seems as if it would break.
Cookie-Less Tracking
A few different methodologies have been used to uniquely identify Internet users without utilizing a cookie. One such method is using the combination of IP address and “browser header” information on the user, which almost uniquely identifies a user. Another method is using the flash registry information that is secretly written in flash plug-ins installed on a user’s browser.
There could be other methods like these that either secretly write something on a user’s computer without his permission, or use a characteristic of this system to uniquely identify him without his permission. These methods typically won’t give the user an option to remove/opt out of such tracking. Most of these types of tracking methods can be integrated with Personally Identifiable Information (PII) data, which can then be used to target the Internet user with messaging without his permission and without giving any option to opt out.
The problem with adopting such methods is that they fall within a gray area of legality as far as Federal Communications Commission regulations are concerned. There is a reason why all established advertising technology companies use cookies regardless of the fact that they sometimes get deleted.
A final point about these methods is that they are much less accurate than using cookies. For example, there is one such tracking company with a demo on its website showing how well its technology can track you at a given time (at the time of the demo), yet your signature changes every time you make changes to your browser or you get a new IP address from your DHCP service. As a result, this method is changing more frequently than cookie deletion, making the data less reliable.
Critical Mass
When it comes to cookies, like anything associated with analytics, it’s purely a numbers game. The fact is that when algorithmic attribution is performed on all the touchpoints associated with all the users who were exposed to your digital assets -- both converters and non-converters -- over a given timeframe, the universe of touchpoints is easily in the millions, often in the billions, and can even occasionally top the trillion mark. So when a percentage of users delete their cookies at some point during their path to conversion, it definitely adds a slight level of inaccuracy to the attribution equation, but given the critical mass of the overall sample size, the degree of inaccuracy created by cookie deletion is miniscule. Using statistical and scientific methods, an intelligent attribution management system can accurately distribute credits using cookie data.
Sweat The Things That Matter Most
Instead of sweating the few missing touchpoints in your data that can easily be compensated for by scientific methods, focus on factors within your attribution initiative that matter the most, such as how much your brand equity affects conversions; how much impact recency, frequency, lag time and the length of your sales cycle have on performance; or whether exogenous factors as the economy, weather, or world events affect your marketing ecosystem. Use your attribution solution to answer these larger, more important questions instead of sweating the small stuff.
Now I know why they call them cookies, because everyone is fighting over crumbs.
Talk about a bridge to nowhere.
What are cookie deletion rates?
Also you make some aggressive statements in the cookieless tracking section. Quite frankly it seems to be riddled with misinformation and 0 sourcing for your broad generalizations. I would suggest at least imbedding a few links next time to add credibility to your argument. Quite frankly I would expect a more buttoned down post from someone who uses an algorithmic approach to anything.
I agree with Nate above. Very aggressive statements against cookieless tracking with information included to intentionally mis-inform. There is one cookieless tracking provider that is completely privacy compliant and does not collect PII. It is interesting that you attack clear differentiating functionality that exposes most attribution providers’ inability to track cross-browser/cross-device. Which I am sure you would agree, is very valuable. Your assessment of cookieless tracking being less accurate due to the changing elements only holds true for vendors that use exact match tracking (typically used in fighting online fraud). Using methods that expect your “signature” to change over time actually produces very useful data. Finally, Sweat the things that matter most? How can you even figure out the length of your sales cycle and lag time unless you have tracking that can see back that far? Cookies are plagued by the in ability to consistently look back farther than 30-45 days.
@Jon Shoff and @Nate Carter @Mike Einstein
Thanks to Jon, Nate and Mike for your recent comments. One of the challenges of writing this monthly column is having to write to an audience whose knowledge of the analytics industry ranges from novice to advanced, so content may at times assume a background understanding of a topic that may not be held by some readers. That being the case, readers can gain a great understanding of the use of cookies by reading this article written by Google’s Avinash Kaushik: http://www.kaushik.net/avinash/web-analytics-visitor-tracking-cookies/.
To simplify the point about tag-less methodologies, in these days of heighted consumer awareness of privacy, and government’s increased regulation of privacy issues, online marketers need to ask themselves several very important questions. 1. Can you say that your ad tracking system uniquely identifies a user only if the user has provided you with permission? 2. If you have the user’s permission, do you provide that user an option to stop or opt-out from getting identified later? 3. If you have that user’s permission, is it possible for the user to delete the information that your ad tracking system has written on the user’s computer? If you answer is NO to any one of these, then your tracking practices fall within a legally gray area that may leave your brand at risk.
When I saw the title of this article, I had high hopes that it might include some insights into what % of cookies are being blocked or deleted, and how to quantify or account for the impact of data loss when reporting results.
Unfortunately it didn't address either. Rather, it seems to be a manifesto on "please stop talking about cookie deletion!" Was that the point?
FWIW, an interesting piece, and a valid point of view, despite what the other commenters have 'added'. Some of the major publishers already have algorithms in place for 'persistent IDs' to identify users anonymously even after cookie deletion; whether it takes off on a bigger scale is harder to tell, due to the legal themes you identify.
As for cookie deletion rates (again, directed at the commentards), there's this amazing tool, that allows you to search for things on the internet. It's called a 'search engine'. You should try it - you'll rapidly find that comScore suggests cookie deletion rates of c30% (and that not everyone agrees with that figure).
I think Anto makes a fair comment that given the large dataset size of both converters and non-converters when doing Attribution Modelling means that bias induced by cookie deletion from a relatively small sample is unlikely to cause any significant errors in statistical output. The reasoning however assumes that the default option used by websites is an Opt-in meaning that by default users are opted in and tracked and then they have the option of opting out. I am not sure how tenable this position is with the arrival of the new ePrivacy Regulations - especially here in Europe - which means that there is a strong focus on having backup plans of cookie less tracking. This is not sweating on small stuff by any means!
I think the focus on cookie deletion rates in the comments here are a bit misplaced. There isn't a simple rate as cookies are deleted over time. The real impact of cookie deletion needs to consider both the time that has elapsed. When when attributing to multiple touches, deletion has a bigger impact on early touches. That has a real impact on how we need to view cookie deletion.
There are some good anecdotal stats you can infer on the impact of cookie deletion on some types of reporting through Google Ad Planner as well. I have rolled up a few examples in a post here (http://b2bdigital.net/2012/04/16/cookie-deletion-and-measurement-what-you-need-to-know/) as well as some of the implications it has on marketers. The really interesting point to me was the wide range of the impact across different sites, it really highlights that you can't just use a simple rule of thumb here.
Best,
Eric