Analytics Roundup – Updates from September 2024

In my monthly roundups, I cover recent articles and items of interest, mostly related to the Google analytics stack, including GA4, GTM, Looker Studio, BigQuery, and CoLab. Click here to subscribe.

Observations and Musings

Here are couple of recent themes I’ve noticed in conversations with the analytics community and clients:

1. People’s relationship with GA4 is transitioning from “is it me?” to “it’s not me, it’s you.” A year ago, I got a lot of how-to questions from people who felt lost in GA4. Now, I’m fielding a lot of questions about why data is wrong. The problem with GA4 is that there is so much squirreliness inherent in the metrics that it can be hard to know when you actually have an issue that can be fixed. The good news is that many of the problems I’m encountering do have solutions. Common themes are bots, consent and website redirects.
2. Speaking of bots, there has definitely been an increase in form spam recently. I suspect the increase is due to Google’s push on Performance Max campaigns: PMax = more display advertising, and more display advertising = more spurious conversions (or Key events, whatever). Why? Because Google and every other display network are greedy for inventory, so they end up doing deals with shady network partners that cook the metrics in order to get served more ads. The spam submits range from legit humans entering bogus results to bots submitting forms without even visiting your site, so there is no great way to prevent it from happening, but you can verify legitimate submits with enhanced conversions in Google Ads, the conversions API in Meta, and similar features in other platforms.

Another thing that is consuming me right now is the proliferation of resources and advice related to AI. I love learning new things and I’m generally pretty good at it, but right now I’m on overload. My Medium feed has new articles every day claiming that some new architecture is better than everything that has come before. I’ve also incorporated AI models and tools into my day to day work, but the landscape is changing so fast that every project I take on feels like a do-over. I don’t subscribe to the singularity hypothesis, but it sure seems like we are approaching an inflection point. 

Naturally, I asked my most trusted advisor for help…

(Okay, fessing up: I might have made up that answer. But I totally know that’s what Claude was thinking.)

The New GA4 Transactions Report

Google added a report that shows purchases and revenue by transaction ID. It should appear automatically in reporting navigation under ‘Monetization’, but if you don’t see it, you can find and add it in the report library.

The main benefit of the report from my POV is for reconciling GA purchase data with your ecommerce platform. Towards that end, I recommend customizing the report and adding Shipping amount and Tax amount.

On the subject of new stuff in GA4, I talked about it last month but it took a bit for the plot rows feature to roll out to all GA4 properties. Google also added anomaly highlighting to standard line chart reports. I’ve used both of these features several times in the past few weeks to troubleshoot and/or analyze data. It pains me to say anything positive about GA4 reporting, but I really appreciate both of these additions.

Funnel Charts in Looker Studio

Yay, Google just added a funnel chart option to Looker Studio! As is, they work with a combination of a dimension and a metric that naturally form a funnel. In my opinion, the most obvious use case for this chart is to show sequences of GA4 events, which is not easy to do without a bit of trickery. I wrote a short blog post describing how to do this. I’ll create a video walkthrough soon.

Google Ads Measurement & Enhanced Conversions

Enhanced conversions

In this video, Dana DiTomaso speaks with Ginny Marvin of Google about enhanced conversions in Google Ads and GA4. Ginny does a great job of explaining why they matter and how they work, and Dana and the live audience asked most of the questions I had as I watched:

Everything You Need to Know About Google Ads Measurement for 2024 & Beyond with Ginny Marvin

GA4 Google Ads reporting

For a primer on how GA4 Google Ads reporting works, check out this video from Benjamin Mangold:

Google Ads Reports in Google Analytics | Standard, Exploration & Advertising

If third-party domain is part of your checkout, Google Ads tracking may get lost

Related to this, I recently spent a while chasing down a Google Ads / GA4 purchase conversion discrepancy issue and discovered something interesting: if your checkout process kicks a user out to a third party somewhere before order confirmation, and you don’t have the third-party domain on your Referral exclusion list, GA4 will correctly attribute the purchase to source / medium = google / cpc, but the purchase may lose the connection to the Google Ads GCLID. The GCLID is necessary for importing the purchase into Google Ads, so you will end up underreporting purchases in Google Ads. This only matters if you are optimizing on GA4 conversions in Google Ads, but for one client it was underreporting by about 80%!

The issue can be really hard to detect if your GA4 property isn’t tracking on the third party, because GA4 doesn’t restart a session if a user arrives via a third-party domain mid-session, so there can be very little evidence of the third-party domain in GA4 reporting. This is pretty convoluted to explain, so it’s probably hard to follow what I’m saying, but the main point is: if you use a third-party payment processor, booking system, donation platform, etc., make sure to include its domain in your referral exclusion list. If you have your GA4 tag on the third party, also make sure to add it to your configured domains.

Dynamic Metrics Selection in Looker Studio

This isn’t new, but it is new to me and I LOVE it! Mehdi Oudjida described a method for selecting which metric is displayed in a chart with a dropdown or fixed control, which is way more intuitive than Looker Studio’s Optional metrics feature. I often want to give the user the ability to change metrics, so this trick is a godsend.

In his article, he shares various other cool hacks and links to an article that demonstrates the same technique with dimensions rather than metrics.

Here is a chart I created using Mehdi’s technique, using Google Search Console data. On the left is the metric selection control. You can change the metric displayed in the chart by clicking on the menu.

Content We’ve Published

• Documenting a Google Tag Manager Container
  Last month, I described how I used ChatGPT to speed up creating Tag Manager documentation. That process worked pretty well, but I ended up building a tool that works even better.
  Bonus share: along the way, I came across Robin Lord’s Tag Mapper tool, which produces killer visualizations for understanding and communicating GTM issues and complexity.
• Funnel Reporting in Looker Studio with GA4 Events
  How to use Looker Studio’s new funnel chart to visualize sequences of GA4 events.

Articles/Videos That Made Me Smarter

• The Counterintuitive Birthday Paradox
  I’ve used the birthday paradox many times as an example of why we can’t trust our instincts when it comes to probability, but if I’m being honest, I never really understood the mathematical principles behind it. I finally feel like I get it, thanks to Shri Khalpada’s brilliant explanation and visuals.
• Unlocking Consumer Insights in the $144 Billion Online Education Space
  I’m a big believer in the business insights search query data can provide. Jonas Sickler demonstrates this beautifully in his analysis of online education. It is worth checking out for the insane “online college”/unemployment correlation alone.
• When to call an A/B test if your results aren’t clear
  I really enjoyed this piece by Amanda Berg. In it, she explores the gray area when data doesn’t provide a definitive answer, but does have directional value. IMO, this is where marketing data analysts live 24/7. Even if the volume of data we have meets statistical significance, we have to contend with sampling bias caused by consent requirements, device-based user tracking, etc.
  She also shares a really nice significance testing tool 🙂