Anaytics Roundup – Updates from March 2025

I feel stupidly cliche saying it, but the thing that I’m most excited about right now is AI – I’ve been using machine learning methods for data analysis for a while, and have been caught up in the LLM hype like everyone else, but I feel like I had a bit of a breakthrough this month. 

It happened when my team built a custom GPT for analytics troubleshooting, affectionately named “The Debuggernaut”. It was dead-simple to do, and it really works! Our goal was to create something to provide a framework for troubleshooting and a structured way of communicating our work to clients, and that’s exactly what it does. And it’s going to keep getting better as we feed it more useful and relevant information to draw from.

We were inspired by my colleague Kris, who built a GPT for HR documentation (talk about something you want to automate 🙂). Then I read an article by Seer Interactive on a GPT they built (see below). I also found a prompt template Rick Dronkers created to improve BigQuery SQL results (also below). I hadn’t previously appreciated how much better an LLM can be with just a little help.

(Note that if you have a Claude subscription, Claude Projects provides a similar capability.)

Also below you’ll find tools and techniques for tracking if and how a website is getting cited by AI tools, the late but welcome addition of annotations to GA4, and a new GA4 BigQuery transfer service.

Product Updates

• Finally, annotations in GA4!
  When GA4 was first introduced, I really missed this feature from Universal Analytics. Now, I’m not sure how much I’ll use it, since I prefer Looker Studio to built-in GA4 reporting. Still, they did a really nice job with it. In particular, system-generated annotations could be really helpful.
• BigQuery has added a transfer service for GA4 reporting data
  This is summary reporting data, in contrast with the GA4 BigQuery export, which is very raw, event-level data. On the plus side, the transfer service data can be used as-is for reporting. On the minus side, it’s very limited. Read my full writeup here.
• Google Tag Manager will start auto-loading Google Tags for Google Ads and CM in April. The TL;DR is: it’s fine to do nothing.
  Simo Ahava wrote this article explaining the change: Clarification On GTM Auto-Loading Google Tag For Ads And Floodlight Events
  If you find Google Tags a bit mysterious, and worry about things like duplication or combining tags, I highly recommend reading the comments at the bottom of the article. It’s basically Simo doing a Q&A on Google Tags.

Workflow

• Project Brief GPT: Advancing AI-Driven Workflows at Seer, Jordan Strauss, Seer Interactive
  Per above, this article helped me think about how we can be using GPTs better to improve processes in our organization.
• Quick Tip: BigQuery LLM Prompt, Rick Dronkers
  A handy template for improving LLM results when asking for BigQuery help. I really like the  <google_analytics_4_best_practices> section of the prompt in particular. If you’ve tried ChatGPT or Claude for generating BigQuery/GA4 queries, you know they tend to fall short. This template results in big improvements.
• BigQuery’s Ridiculous Pricing Model Cost Us $10,000 in Just 22 Seconds!!!, Yingjun Wu
  Yikes. I’ve been pretty diligent about optimizing query costs, but this gives me shivers. Time to get some better guardrails in place.

Attribution

• How to Track Traffic from AI Overviews, Featured Snippets, or People Also Ask Results in Google Analytics 4, Dana DiTomaso
  If you want a more nuanced view of how different types of Google organic results drive value on your site, this is a very clever solution.
• Tracking visibility in LLMs – on a related note, there was a lively discussion on the SEO Community Slack group on this topic. Several people recommended Knowatoa. I tried out the free version – the UI is nice and it seems like it could be a really helpful tool, but it made me want the paid version ($39-$79/month). Other tools mentioned were: Peec, Semrush, Otterly.ai, Profound, RankBee, Waikay.io, and Essio.ai.

Ideas

• The A/B Testing Industry: Why Attitudes Matter, Matt Gershoff, Conductrics
  Ostensibly, this article is about A/B testing methodologies, but it’s really a trojan horse containing a beautifully simple exposition of the allure and cost of complexity. Paragraph four (“Ironically, this argument…”) really landed for me in particular .
• you can just do stuff, Alice Maz
  A nice kick in the pants. 

Privacy

• (data) permanence, Kaiser Fung
  A chilling warning that personal data lives on, even when you request to have it deleted.
• Data Privacy Laws in 2025: Current State & New Developments, Phil Pearce, MeasureMinds
  Handy reference for privacy regulations around the world.
• Microsoft Advertising will start enforcing Consent Mode in May, Anu Adegbola, Search Engine Land
  This capability enables UET (Universal Event Tracking) and conversion tracking when a user declines consent, minus any user-identifiers in the data. It will only be enforced for European advertisers, and is a Microsoft requirement, not a regulatory requirement – as long as you have a consent solution in place that complies with local laws, you are fine ignoring it from a legal standpoint. If you do decide to set it up, here’s a handy guide.
  I really, really hope platform-specific consent protocols don’t become more of a trend. 

Miscellaneous

• dataLayer.push: What is it? (with examples), Julius Federovicius
  I often find myself explaining the dataLayer to developers. I’ve bookmarked this to save myself time in the future.