DATA DRIVEN ATTRIBUTION CONFUSION IN GA4
GA4 has made the data driven attribution model available for all accounts – something that was only available for GA360 accounts during the Universal Analytics days. The data driven attribution model – or DDA – is arguably (keyword) the best model when it comes to attributing a conversion to channel(s) that drove that conversion, because it looks at various factors like number of touchpoints, recency, incrementality etc.
The version of DDA that GA4 uses is also different, in a good way, when compared to the one that Universal Analytics (UA) used. For example, the new DDA model looks at up to 50 touchpoints compared to a mere 4 in UA. The older model was quite restrictive in the sense that you had to have a minimum of 400 conversions per 28-day period for the DDA to work but in GA4, the only requirement is that you have at least one conversion set up.
There are other differences as well, but we’ll talk about those another time. The above is all well and good but where the heck is this new DDA in GA4?
GA4 documentation suggests that the standard reports use the attribution model selected under the Attribution Settings section below –
This model is set to DDA by default. A great feature worth highlighting here is that you can chose between last interaction and DDA models from this section and conversion data (including historical) changes with the attribution model, without impacting anything else.
To the inexperienced user who hasn’t gone through the Souls games style frustrations of using GA4, having the attribution model set to DDA means that all reports in the platform use this model. But they’d be wrong because most reports in GA4 use the last interaction model! This is because the DDA is only available for reports that use event-scoped traffic dimensions (more on this later).
So Many Dimensions
Before we talk about reports that use the DDA, let’s talk about one of the most confusing aspects of GA4 – if you’ve been using the platform for a while, you’ll have noticed that there are three different sets of traffic dimensions available – ones starting with first user, like first user source, another set starting with session, and a third set without first user or session, similar to what UA had.
The first user set is available under the User acquisition report –
These dimensions are attached to the first_visit / first_user event that is used to calculate the new users metric for a GA4 property. This means that if a user comes to your website via Organic for the first time, that user will always fall under Organic even if they use another channel for subsequent visits to the website – they are locked to the first channel used to enter the website. You’ll notice that this report doesn’t provide total or active users metric and only shows new users – this is because that’s the primary metric that the first user set of dimensions are meant to show. The conversions shown in this report use the last interaction model because any subsequent visit is almost disregarded by this group of dimensions, you can call it first interaction or only interaction – whatever you fancy.
Moving on to the session level traffic dimensions which are available under the traffic acquisition report –
These dimensions are designed to work in almost (I say almost because the definition of a session is different in GA4) the same way as the equivalent dimensions in UA. You’ll notice that the conversion column looks completely different because the conversions are attributed based on the last interaction model – as Universal Analytics would have. For example, if a user came to the website using Organic first and then via Paid Search to convert, the conversion will be attributed to Paid Search under this report. You’ll also notice that this report shows the total number of users, which wasn’t available in the user acquisition report.
The third set of traffic dimensions is available under the model comparison report –
This is one report which uses the DDA – as the report suggests. These dimensions are all event-scoped and the DDA model is designed to work at the event level and not at the session and user level, which is the scope of the session and first user level dimensions.
GA4 uses an event-based model meaning the lowest unit of measurement is an event and the DDA model is a complicated algorithm designed to look at data at this lowest level of measurement. This is the primary reason why it’s not available across most reports in GA4 – the DDA is incompatible with dimensions that aren’t event-scoped.
Scope It Out
Understanding scopes can be difficult but it’s not a new concept and was an integral part of Universal Analytics. Not all metrics and dimensions can be combined with each other to produce reports because there might be incompatibility at the scope level. Check out Google’s Help Center to learn more about scopes in GA4.
The only other couple of reports that use the DDA model are –
- Conversions – if you click on a particular conversion type in this report, it opens a screen like this –
- Explorations – if you’re using event-scoped dimensions like the ones shown in the model explorer report screenshot above. A useful feature of explorations is that it greys out dimensions and metrics that aren’t compatible with your selection. This is probably the best place to look at your conversion data attributed using the DDA model. An example below where you can look at your device data with event scoped source / medium attributed using DDA.
All other reports like landing page, device etc. seem to use the last interaction model for attributing conversions as far as we’ve tested.
A useful identifier is that the DDA model’s output is usually in a float format (a number with decimals). This isn’t always the case but if you have a significant number of conversions, you are likely to see conversions with decimals attributed to your dimensions. If you don’t see that in your reports, most likely DDA isn’t being used.
Conclusion
Besides the pounding headache, I do believe that the availability of the DDA model in GA4 is a good thing. It at least provides analysts with an opportunity to understand the true value of the channels in market. Google could make it easier for users to figure out which reports use which attribution model – like a big banner up top. In the meantime, be sure to check what attribution model your conversion data is using.