Recent articles and items of interest, mostly related to the Google analytics stack (GA4, GTM, Looker (Studio), BigQuery, CoLab), from Two Octobers’ Head of Analytics.
These have not been announced on What’s new in Google Analytics, but they are documented by Google and several users have spotted them in the wild. I checked a number of GA4 properties I have access to, and don’t see them yet, but here’s hoping we all get access to them soon. (update: Google officially announced the plot rows feature, plus some other interesting report updates on September 3rd – more on these features in next month’s update!)
If you find consent management confusing and anxiety-provoking, at least be comforted that you are not alone. We do a fair bit of work in this area, and there are lots of moving parts, any of which can be broken, and the landscape is changing quickly. I wish I could point you to a resource that explains it all clearly, but I don’t know of one. If you do, please let me know.
A Google Tag Manager container that has been around for a while inevitably accumulates tags, triggers and variables that are not working, out of date or just plain not being used. George Clements shared this very clever solution for quickly identifying and removing triggers, variables and templates that are not in use. Before trying it, I recommend going through your tags, and removing any that are obsolete (e.g. Universal Analytics) or out-of-date (e.g. Facebook or Floodlight tags placed by an agency you no longer work with).
Related to that, I recently found myself needing to document GTM tags, triggers and variables for a client. In a moment of inspiration, I uploaded the Google Tag Manager export into ChatGPT and gave it the following prompt:
“Can you produce detailed documentation of tags, triggers and variables that are set up in Google Tag Manager based on the uploaded export of the Tag Manager container? Please list out each tag, trigger and custom variable. When listing Firing Triggers, please use the name of the trigger instead of the ID. The tag firingTriggerID matches the trigger triggerID.”
I had to proofread what ChatGPT produced, and fixed a few errors, but it saved me a bunch of time!
I’ve written a few times about all the reasons analytics data doesn’t equal truth. One I have not given enough attention to is ad blockers. Jason Packer did an analysis of ad blockers in 2017 and just published an updated version. Spoiler alert: a LOT more people use ad blockers today than did in 2017.
What to make of this? There’s a lot to be said on the topic of web analytics data completeness and accuracy, but this is what I’ve been thinking about recently:
Mathematicians make the distinction between descriptive and inferential statistics, and use different methods when using one versus the other. For the most part, GA4 uses the methods of descriptive statistics, and presents metrics as facts rather than estimates. At best, this is misleading, since the data it collects is far from complete and the concept of a user is several layers of abstraction from an actual human being. At worst, we waste money on marketing, content and web development because we are making decisions based on flawed assumptions.
We would be much better served if Google was transparent about confidence ranges for estimates and methodological limitations of the data introduced by factors such as thresholding, cookie consent, and ad blockers. The more I learn about the mechanics and practical realities of data collection, the more I realize how much we don’t really know. The good news is that exporting GA4 data to BigQuery creates the opportunity to employ inferential methods of analysis and gives us the ability to quantify some of the unknowns.
Learn how to automatically generate documentation for Google Tag Manager with a Colab-based tool. Simplify…
Updates in Google organic search--algorithm update & AI Overview sources--as well as Google Ads changes…
Learn how to adjust settings on any standard report in GA4 to dive into the…