With the transition to Universal Analytics, Google also seems to perform the calculation of referral traffic based on a more sophisticated variant. This is particularly noticeable in a sudden increase in the sessions in the referral channel but can be quickly leveled by simple home remedies.
Online Shop Operators who observe a rapid increase in sessions in the Acquisitions> All References report after transitioning to Universal Analytics are not unsubstantiated by a huge question mark. Theoretically, the sudden increase in traffic via referring, external domains are possible, but in practice very unlikely. And if the deeper look into the sources then reveals that hosts like paypal.com or sofort.com are significantly responsible for the increase, the confusion is complete.
The curious thing: Hardly any other user will actually have been made aware of the offer of an online shop via payment services such as Paypal, which is why the referring domain in the conversion chain has no business at all. However, the bigger problem is that not only the number of sessions but above all the generated revenue is ultimately assigned to the wrong channel and referrer, which makes a later, channel-based evaluation difficult. Because at the time of the increase in sales via referral traffic, sales in channels like Paid or Organic are drastically decreasing. This means shifting revenue-bearing channels and ultimately misassigning them.
The cause of the problem is, of course, obvious and should have come up to the one or the other readers: In checkout pay services such as Paypal or Soft transfer short term to their own domain to complete the payment process. Everyone who has ever made a purchase for such a service knows this. When switching back to the website of the online shop, Universal Analytics finally starts a new session and assigns the generated revenue to the last referrer, because it was tracked just before a transaction as a reference. The resulting problem is explained above, in short: the referrers steal the previous channels quasi the sales. Google itself discusses the phenomenon as follows:
Analytics automatically identifies the last location in front of your site from which the traffic came. The domain names of these websites are indicated in the reports as sources of referral traffic. […] Traffic from a third-party shopping cart […] By default, a reference automatically triggers a new session.
Solution 1: Reference exclusion list
The simplest and most effective solution is to exclude the referring hosts responsible for the sales shift. Of course, it is necessary to identify them first. I would recommend first setting the start date of the period in Analytics to a few days BEFORE the revenue growth to understand the increase. In the list of strongest referrals ( Channels> Referral, Sort by Revenue & Sessions) you should manually click through the first 10-15 referrers and just see if the traffic & revenue at the time of transition to Universal Analytics at certain Hosts has increased enormously (it is best to superimpose both metrics in the graph on top of each other).
Once you have identified the biggest person in charge, you continue with the fine adjustment. For this, you set the time period more precisely, with the start date, in which the high referral turnover started. In order to get the clearest possible picture of which referrers actually provide payment services, one can, for example, define a RegExp via the text filter which filters characteristic terms such as “pay”, “bank”, “visa” or the like from the source:
Once you have checked the identified links again manually (after all, you do not want to exclude any real references) you should at least have already identified a large part of the shifted traffic. In the ideal case, one now compares the actual total unnaturally grown referral traffic and revenue, with the exact traffic and revenue that came from just analyzed hosts.
If the values approximate, go to the Admin section in the last step: Administration> Property> Tracking Information> Referral Exclusion List, and enter all the hosts identified in the last step and for has declared irrelevant to have to be listed as actual references. The implication of this approach is that traffic from all these hosts will not trigger new sessions and will keep existing sessions in Analytics.
Google itself explains this as follows:
You can prevent certain domains in your Analytics reports from being identified as sources of referral traffic. This feature is often used to exclude traffic from a third-party shopping cart, preventing it from being counted as a new session and as a reference whenever customers return to your order confirmation page after checking out on a third-party website. […] If a referral source is excluded, the traffic coming from the excluded domain to your site does not trigger a new session. Therefore, if traffic from a particular site is to be considered a new session, you should not exclude referral traffic from that domain. […]
Beware: If you manually exclude various references in the old Analytics Code, you will have to re-enter them using the new reference exclusion list anyway.
Solution 2: Channel Grouping
The only potential disadvantage of the referral exclusion list is that it can no longer be verified by Analytics which pay services generate the most revenue (or are used in the checkout). This should ideally be understood by other means, but who would like to use these Analytics, the referral exclusion list should not use. One way to still isolate the traffic is to define in the channel grouping a separate channel eg only for payment services. In the Admin area under the item Manage> View data (select)> Channel Settings> Channel Grouping, selecting Default Channel Grouping and defining a New Channel that includes all referral sources of the hosts identified in the last step:
The disadvantage of this variant is, of course, that the actual channels on which eg a transaction was made initially remain veiled. The problem explained at the beginning remains in principle therefore, the references are only sorted differently. In the tool to model comparison in Analytics ( assignment ), however, the attribution can be adjusted, thus the defined now Channel “payment services” for example, simply no value can be assigned to a position based model. This correctly allocates the revenue generated by “payment services” to the previous channels.
The exclusion of references that are not relevant within the conversion chain has first and foremost the advantage of being able to carry out the channel-based e-commerce evaluation in an unadulterated manner. Why the need to exclude payment services has only been observed several times since Universal Analytics is unclear. It is possible, however, that Google Analytics will trigger a new session with the new tracking code (actually correct) and all those referrers who were the last point of contact shortly before the transaction was completed. What’s your experience with tracking pay services in Google Anlytics, and what are the best solutions? I am glad about every comment!