CJ has been made aware of significant changes in the revenue assigned to various channels within Google Analytics (“GA”), caused by Google’s planned migration to Google Analytics 4 (“GA4”).
In some cases, clients report that GA4 is significantly under-reporting revenue for non-Google channels such as affiliate, compared to their previous Google Analytics version. These errors seem specific to GA4 and/or Google’s Data-Driven Attribution (DDA) model. We’re not seeing these changes within CJ tracking or reporting, which remains accurate and stable.
We understand that Google has not acknowledged this issue, so we created the following guide to help our clients navigate internal conversations around GA4. While a migration to Google Analytics 4 should not have any effect on CJ’s ability to track or report publisher value, we understand that it’s important to explain any significant shifts in Google Analytics to the rest of your organization.
This is an emerging issue, to which Google has not yet responded. As such, this guide is based on CJ’s best understanding of the issue. We will update as Google provides more information.
Within this guide, we link to Google documentation wherever possible to help you access their relevant guidance; however, Google’s documentation relating to GA4 is not comprehensive. In some cases, we may point to non-GA4 documentation that we believe is still relevant.
Why This Shift is Happening
It seems that a notable number of the issues are attributable to Google’s significant expansion of their Data-Driven Attribution (DDA) model in GA4, and the issues attached to that model when applied to non-Google channels. We explore these issues in the guide below:
If you were not using Data-Driven Attribution in your previous GA version
- In previous versions of Google Analytics, most clients were set up to report on a Last Click basis. GA4 instead defaults all accounts to the DDA algorithmic attribution model.
- DDA is not a last-click model, so it will show significantly different channel performance from previous last-click setups. DDA also has significant issues measuring the touchpoints and value of non-Google channels (including email, social, and/or affiliate).
If you were using Data-Driven Attribution in your previous GA version
- Google’s DDA weighting changed drastically in GA4. Their algorithmic model was changed from considering 4 touchpoints, to 50, and increases the measurement of touchpoints in the journey which typically favor display and search.
- This significantly amplifies the extent to which DDA manipulates your tracked conversion data, as well as the extent to which DDA’s known issues affect your non-Google channel attribution.
A Guide to Google Data-Driven Attribution & Non-Google Channels
DDA Was Designed to Measure Google-Owned Channels
DDA documentation states that it is designed to “look at all the interactions…. on your Search, Youtube, Display and Discovery Ads”. As currently designed, DDA does not accurately track or attribute revenue for non-Google channels, including email, search, or affiliate.
This means there are significant issues that arise within DDA when applied to some or all non-Google channels. This could result in large data gaps and inaccuracies in DDA’s revenue attribution.
Google terminology used in this section:
- A conversion type is a category of conversions—for instance, an “ecommerce transaction”, “phone order”, or “app download”.
- A conversion path is the journey a consumer takes to conversion—for instance, a user who sees a display ad, clicks an affiliate link, then converts (display > affiliate > conversion). In Google Analytics 4, conversion paths can contain up to 50 touchpoints, and the same touchpoint can be counted multiple times. For instance, users who see two display ads (display>display) are considered to have taken a different conversion path than someone who sees three display ads (display > display > display).
Issue 1: Google Acknowledges Potential Gaps in DDA Methodology
Google states in their documentation that there are common scenarios where they cannot provide measurement for valid data. This particularly applies to gathering crucial data on conversion paths.
DDA compares different conversion paths to infer the value of the channels within that path. To do so, DDA requires at least 400 conversions per unique path over 30 days. If this threshold isn’t met, all journeys which took that path are discarded from DDA. Google acknowledge that these journeys are valid, but that their model is unable to measure them (see “Insufficient data to generate a model”).
Put simply, GA4 discards these journeys because the DDA model is not able to reach statistical significance for that conversion path. Because Google increased the number of possible touchpoints so significantly in Google Analytics 4, the number of possible unique paths increased exponentially, and the likelihood of any path leading to 400+ conversions is much lower vs previous versions of GA.
CJ looked at our data for over 1,200 clients. Google’s methodology would likely exclude ~98% of affiliate conversion paths from its DDA model. CJ observed that only 1.4% of conversion paths containing an affiliate touchpoint resulted in 400+ conversions.
When DDA rejects these valid journeys, they are no longer measured using the DDA methodology. Instead, Google uses “aggregate data” to make assumptions about the value of those interactions, and does not share how these assumptions are made.
Conversion paths from high-volume channels (i.e., Google-owned channels such as display and search) are more likely to pass this conversion threshold, and therefore their journeys are less likely to be ignored by DDA.
Issue 2: Interactions with Google Channels May Be Given Preference in GA4
Google-owned channel interactions are more likely to be assigned value in DDA. Google collects more data on ad interactions for Google-owned channels, and they provide features that measure Google interactions with more consistency and preference than non-Google channels.
Google Impression & Video Measurement
While DDA includes impression and video view data from its own channels, Google ignores all impression and video data from non-Google channels.
The inclusion of Google-only impression and video data creates far more touchpoints for Google-owned channels than non-Google channels. This creates significant potential for bias in DDA results.
Google Click Measurement
Google clicks are recorded using Google Click Identifier (GCLID), a unique parameter that is only available for Google channels and can be ‘auto-tagged’ onto any Google-owned links.
Google describes auto-tagged clicks as providing “more detailed reporting” and “eliminating errors”. However, Google requires non-Google channels to use their less reliable “manual tagging” solution (commonly known as UTM Parameters) for identifying clicks in Google Analytics 4. This design choice means that GA4 tracks Google-channel clicks more reliably, and with more detail, than other channel clicks.
When a user click occurs where “gclid” and UTM parameters are present, GA4 ignores the UTM parameters and favors the “gclid”. This means that when there is ambiguity around whether a non-Google channel or a Google channel drove an interaction, the Google channel will always win.
In prior versions of Google Analytics, the option existed to switch this auto-tagging bias off and ensure Google counts non-Google click data at all times (see “You need to use auto-tagging for non-Analytics purposes”). They acknowledged that this option was required to “avoid data discrepancies” in their Analytics products. Google removed this crucial option in Google Analytics 4.
CJ’s Viewpoint: How Does This Apply to Your Affiliate Program?
Google’s current design for GA4/DDA contains inherent and unavoidable biases toward Google-owned channels. Google’s own documentation alludes to such biases. CJ’s data indicates that DDA is not able to provide reliable measurement on the affiliate channel.
With the current methodology, you should not use Google’s Data-Driven Attribution model to measure the affiliate channel, particularly if you are using the updated version of DDA in GA4.
CJ has robust tools available for measuring the value of your affiliate channel, including Cross-Channel Customer Journey reporting, Channel Incrementality tools, Competitive Benchmarking, and more. These tools are available in-platform and can be used to reliably derive the value of the channel, your partners, and your strategies.
GA4 Best Practice: Alignment of Methodology
For clients looking to understand a deduplicated view of all channels’ performance using GA4, last click is currently the most reliable methodology provided by Google.
If your previous version of Google Analytics was set to Last Click:
Ensure your current GA4 settings are also set to Last Click. To access this setting, log in to Google Analytics and go to Admin > Attribution Settings, and adjust your attribution model and conversion window to “last click”.
While GA4’s Last Click view is still subject to mismeasurement of non-Google clicks (see “Google Click Measurement” above), the last-click attribution method will be more reliable and less prone to misreporting than DDA.
If you were using DDA in your previous version of GA:
There is no current method to align with Google’s prior methodology. An apples-to-apples comparison to your previous DDA numbers is not possible. Google’s changes have caused significant discrepancies in reported numbers between the two versions.
CJ will be releasing a further guide for best practices when implementing GA4/DDA as more data becomes available. We do not yet have a position on whether it is possible to overcome DDA’s limitations enough to accurately measure non-Google channels.