.png)
When GA4 was launched, there was a lot of excitement about the ability to change the default attribution model for the platform and finally move away from a last-click world. In reality, this is not how the platform works - there are actually several different attribution models used across the platform, and data using your ‘default model’ is unlikely to be what you’re looking at.
For many years, Universal Analytics has been an important source of truth for marketers - a standard deduplicated way to view performance from several different platforms. With the move to GA4, for many retailers we are seeing big shifts in the way our data looks.
This means that the data we previously trusted is being interrogated in ways that it never has been before, and activity which has been run successfully for many years is being perceived to have stopped working or be working much more effectively entirely due to changes in the way the data is presented. In this blog, we attempt to explain what some of these changes mean.
Session Based Reporting
The main attribution model used in Universal Analytics was “last non-direct”. This means that any conversions a user completed on a site were attributed to the last non-direct channel they interacted with.
If you want to see data in a format similar to what you would have seen in UA, the usual report you’d look at is the “Traffic acquisition: Session default channel group” view in the Reports tab. The attribution model used in this tab is also “last non-direct”, just like in UA!
But wait… if it’s the same attribution model, why does your data look different than it looked in UA? If you’re seeing shifts in session volumes or which channels are driving the bulk of your revenue, read on.
How is a session defined?
An oft-overlooked but very important difference between UA and GA4 is how sessions are defined. A session will still end when there has been a 30 minute period of inactivity, but unlike in UA, a session can now include multiple visits to the site with different campaign parameters, as long as these are within a single browsing time-period.
This means that a person could access the site through a PPC campaign, from an email and through a voucher site and these all be included within the same session if they were in a close enough time period.
If there are multiple clicks through to the site within a session, the first parameters of the session are the ones which define it.
What are the implications of this?
Sessions may be down: The first implication of this is that the number of sessions overall will be lower as visits which previously would have been defined as multiple sessions will be consolidated into fewer sessions.
Bottom of funnel channel revenue may be down: The next implication is that in sessions ending in a conversion, where there are multiple source/medium parameters, it is the first of these which will be attributed the conversion. So the model assigns the sale to the first source of the last non-direct session.
In this example, the conversion would be attributed to the Google Shopping ad click as it is the first source of the second session
Examples of how this affects reporting
Google shopping behaviour
In Google Shopping campaigns, a typical behaviour is that users might click multiple shopping links within a browsing session as they compare the different products that appeared in the Product Listings.
These clicks will be measured separately in the Google Ads or Microsoft Ads platforms, but will be counted as one session in GA4 if they happen in a close enough time frame, so sessions reported may be lower than the number of clicks for the same campaign.
Voucher codes
The biggest revenue driving activities in an affiliate program will typically include bottom of the funnel revenue-driving activity from voucher code and cashback sites. Deal-hunting customers will typically find products they’re interested in buying and immediately search for voucher codes and cashback as a part of their decision-making process.
This common user behaviour can cause a significant change in attributed revenue to the affiliate channel in GA4 compared to what was previously being seen in Universal Analytics because the click from the affiliate site will be bundled into a session with the first click from another channel source.
In this example from earlier, GA4 session-based reporting attributes the sale to Google Shopping where UA would have attributed to the click from the voucher site
Emails
Similar to the above example, codes and offers in email marketing can mean a drop in revenue attributed to the channel compared to what would have been seen in UA.
Customers may reach the site through one channel, find a product to buy, sign up to emails and click through the email to receive a discount all within the same session. This would have previously been attributed to the email channel but will now be attributed to the first channel source in the session.
Other reports
As mentioned above, there are several different ways to view data within GA4:
How does DDA work?
Data driven attribution is what got us all so excited about GA4 - it compares different paths to conversion and assigns value based on the calculated likelihood of a user converting had each channel not been present in their journey. It takes into account how recently they interacted with a channel, corrects for paths which didn’t result in conversions, and should give us a much better view on the incrementally of our advertising than traditional attribution models.
So what model should I use for reporting?
Unfortunately this change has come with a lot of confusion, to the detriment of certain channels, and makes year-on-year comparisons very difficult. However, armed with the knowledge of why we're seeing the specific changes we're seeing we can learn a lot about the behaviour of our customers and use this to make decisions about how we want to move forward.
The good thing about this new change is that we're moving away from using a single view of performance - last click attribution always benefited conversion-driving channels to the detriment of earlier steps in the customer journey. It's time that we start to work on getting a more accurate representation of the impact different channels drive throughout the funnel.
There's no perfect attribution model or way to look at performance: What's right for one business will be very wrong for another. The attribution model which is easiest for reporting is the session based default channel report and so this is likely the model which is going to be used the most going forward. Armed with knowledge about the limitations of this model, we can make sure we're regularly taking a view of how our data looks using other models so that we're not limiting important conversion-driving activity which has served us well in the past.
