It is critical to track user acquisition source to identify the best performing advertising campaigns. In this tutorial, we explore the Google Analytics source attribution process. In other words, what piece of information is recorded when.
Attribution is about specifying a referral source of a particular activity. Those activities are typically macro-conversions or micro-conversions, macro being things like purchases, micro being things like registration, email sign-up, blog comment, and so on.
Ideally, each time a conversion event occurs, a referral source is recorded. But how is the source determined?
The reality is that users often come from many sources before they hit/commit a micro or macro conversion (for example, they may come to the site via organic, then leave, then come via paid search, then leave, then come directly to the site itself). This source tracking information is often provided to the site via UTM parameters, but there are more sophisticated systems out there too. For our purposes, we will focus on UTM.
When the UTM parameters are specified on the URL, these get parsed out and placed into a Google Analytics cookie. If a website does not have Google Analytics, there is no point in having UTMs. Google Analytics has rules for how it deals with a user who hits multiple URLs with UTMs in the course of their lifetime (more on that later). Assuming the website is configured to capture UTM parameters into an external database, any time a micro or macro conversion happens, whatever is in the Google Analytics cookie at the time of conversion would get replicated to the database.
Last click attribution is the most common attribution model employed by Google Analytics. In this case, the Google Analytics cookie represents the UTM parameters for the last, or most recent, source prior to the conversion event, and this is what is recorded in the database. Note that the Google Analytics cookie only overwrites the previous UTM parameters if the user clicks on a new URL that contains a new set of UTM parameters.
For example, consider a user who first visits a website via Google AnalyticsGoogle AnalyticsGoogle Analytics paid search, then returns via organic search, and finally comes back to the website directly or via an email link without UTM parameters before the conversion event. In this example, the Google Analytics cookie says the user’s source is organic, since this represents the last source prior to the conversion. The path of the user prior to that final conversion event is ignored. If instead the user visited the website from an email link with UTM, then the Google Analytics cookie would say that the source is “email”. Therefore, if there are existing UTM parameters in the cookie, and the user comes in via direct, the Google Analytics cookie will always show the UTM parameters rather than “direct”.
A specific user’s Google Analytics cookie parameters will be erased when the cookie expires, or when a user clears his or her cookies in the browser.*)
Some paid attribution tools will allow you to capture “the pancake stack” of sources in a user’s path. In that situation, in our above example, first click attribution would tell us paid search. Alternatively, a minority of websites implement their own cookies that capture a pancake stack and store the first source into their database.
Google Analytics has some more robust functionality in their web interface that lets you perform four different attribution models: first click, last click, linear (divide revenue equally across all the sources in the path), and weighted (customized attribution).
Now that you understand what is the attribution model for each micro or macro-conversion, the question becomes what do you do with the totality of a user’s conversions? For example, look at the UTMs recorded based on the GA last click logic:
Here is where you ask: How much revenue did I get from paid search? From email? From organic? You could say the answers are 5, 50 and 10 (that is, whatever the last source was), or you could also say that you attribute all revenue to the first source (that is, all 65 goes to organic). You could also apply some weighted analysis or apply the linear model (that is, roughly 22 each).