The impact of GA4 and why it shouldn’t be used to validate affiliate transactions
Since 1 July 2023, when Google made the decision to switch from supporting Universal Analytics (UA) data to GA4, Tradedoubler has seen an increase in advertiser uncertainty around the company’s affiliate network tracking, thanks to GA4’s default Data-driven attribution (DDA) giving very different results.
To help clear up the confusion, Tradedoubler’s advice to clients is to understand how each attribution model works and how the company’s constantly evolving tracking system is an advanced, multi-layered attribution model designed and purpose built for the affiliate industry – last affiliate click wins.
Tradedoubler has seen a direct link between GA4’s new default attribution model and its higher discrepancy with non-proprietary marketing channels, such as affiliates. You can see a breakdown of these findings in our blog: https://www.tradedoubler.com/post/goolge-analytics-4
As an evolution of UA, GA4 gives you the ability to alternate freely between three attribution models – all of which allow you to analyse and attribute conversions to different touchpoints of user journeys.
GA4’s three attribution models are:
Last Interaction: all conversions are credited to the last touchpoint a user took before completing the desired outcome. (Similar logic to UA).
Last Non-direct Click: last interaction, excluding direct traffic. Out of the available options this model is best suited to affiliate marketing.
Data-Driven attribution (Default): this model looks at numerous datapoints (data sampling) to dynamically attribute value to each touchpoint based on its actual contribution to conversions, considering factors such as the sequence of interactions, time lag, and the significance of each touchpoint. The DDA was intended predominantly for measuring Google’s own channels – https://support.google.com/google-ads/answer/6394265?hl=en
Google evolving its product by offering multiple attribution models highlights how data can be looked at in numerous ways. We sat side by side with a top 10 UK client to see the difference in data from the three models – finding a 13.5% increase in affiliate driven order value from the Data-Driven model and the Last Interaction model, and a further 24.7% increase from Last interaction to Last Non-Direct Click. Google are highlighting there is no one source of truth, when reviewing traffic, we advise clients need to ensure they use data sets that makes sense to the use case – attempting to align different conversion / click / order value attribution models, such as DDA and affiliate, does not make sense.
Note: Clients can add “user ID” which is the Google equivalent of cookieless/cross device parameter EXTID which will help with discrepancies https://support.google.com/analytics/answer/9213390?hl=en
Tradedoubler’s Tracking Solution
Tradedoubler’s tracking, is another attribution model, built specifically to track and allocate sales to an affiliate last click. When a client signs up to Tradedoubler, tracking is one of our USPs. Our advanced, multilayer tracking includes features such as cross device, anti-adblock, fingerprinting and more, which is constantly being evolved and worked on by our specialist tracking team. Our tracking is built on predefined logic, like the three attribution models Google offers. Tradedoubler sits side by side with its clients to help them understand why our tracking model can and should be used as their affiliate channel primary source of truth when validating sales.
Points of note:
Sessions are no longer exclusive to a source. The definition of a session has changed between UA and GA4. GA4 entering a site from a different source with another session already started will no longer create another session. Meaning within a session time window (default 30min), a user can visit your site from a Google Search result, from a newsletter and end using a cashback site. Three unique sources would report a click, but GA4 will report one session. Universal Analytics would have counted three sessions.
GA4’s approach lacks emphasis on probabilistic attribution, overlooking the vulnerabilities associated with first-party cookies, as well as challenges posed by increasing adoption of ad-blocking software and tracking prevention mechanisms by browsers. Our data indicates that 60% of transactions rely on cookie-based validation, a trend we expect to persist. However, the rising use of ad blockers and ongoing efforts by browsers to enhance tracking prevention mechanisms pose ongoing challenges. Tradedoubler tracking consistently adapts to these challenges, whereas GA4 does not address them effectively.
GA4 does not store channel information when cookie consent is not granted. The traffic appears as source/medium “direct/none”.
Clients who insist on using DDA as their source of truth are advised to use the “Traffic Acquisition” report where transactions are attributed to source/medium with last click – https://support.google.com/analytics/answer/12923437?hl=en
GA is often blocked by ad blockers, tracking prevention or Safari 17 (ITP 0) – in which Tradedoubler’s tracking is not as susceptible to as we are actively working around the limitations. Our measures to work around tracking blocking tend to last for a long time, as an example our “anti-adblock” alternative domain is still up and not blocked for around 10 years.
The “Look-back window” in GA4 is set by default to 30 days. The Look-back window is the timeframe which GA4 uses to determine the earliest source of traffic that can be included in the attribution of a sale. As Tradedoubler clients can request to have a cookie window of longer than 30 days (often set to 45 days) this can cause naturally occurring discrepancies – affiliate traffic being tracked by Tradedoubler with a 45-day cookie window but validated with GA using 30-day look-back period will not align due to the 15 days of excluded traffic.
Questions?
Do you have any questions? Or would you like to discuss how this benefits your business and how to deal with the issue?
I'm a paragraph. Click here to add your own text and edit me. It's easy.