Analytics Roundup – Updates from July 2025

My roundup of recent analytics news and ideas. I mostly work in the Google analytics stack and I’m a little bit obsessed with the intersection of AI and analytics. 

Playing with the new Google Analytics MCP Server

A few days ago, I set up the Google Analytics MCP Server on my computer and connected to it in a Claude Project. If you are not familiar with MCP, it is a protocol for connecting a language model to an external system. In addition to enabling the transfer of data or instructions, an MCP server also tells the language model what its capabilities are. So, with Google Analytics MCP server installed, I can converse with Google Analytics via Claude. Holy shit.

I then put together project instructions for Claude, providing a framework for my conversations. I describe the whole process here, but the TL;DR version is that it took me a minute or two to generate a detailed analysis of a Google Analytics property, and there were insights in there that were, in fact, insightful! 

My plan is to use this framework to bulk-generate insights, then curate what is most interesting and useful, and work with Claude on further exploration. When budget allows, this will lead to deeper analysis in BigQuery, jupyter, etc., but in many cases we will be able to provide valuable insights at little cost to ourselves and our clients.

I’ve read a lot recently about why the judgement and experience of human analysts is irreplaceable, and I agree, but it’s pretty clear that my little experiment is a harbinger of capabilities that every marketing stakeholder will have at their fingertips a few short months from now. 

Product Updates

I include updates here that I deem significant. For a complete list of recent GA and Looker Studio updates, see Google Analytics Release Notes and Looker Studio release notes.

• Looker Studio: Highlight Charts by Filter – not a major update, but if you’ve ever struggled to figure out which charts have what filters applied, this is very helpful. Here’s hoping they do something similar for data sources!
• Introducing pay per crawl: Enabling content owners to charge AI crawlers for access, Cloudflare
  If they won’t pay Scarlet Johannsen, honestly, you’ve got no hope. But this feature set also gives you control over what AI crawlers can access for free. That part is useful. 
• Google Trends API – SEOs have been asking for this for years. They’ve also normalized the interest metrics in the API, which is a very welcome addition.
  This article from Dan Taylor goes into a bit more detail.

Workflow

• LLM Agent-Based Browser Automation For Digital Analytics — Using VS Code & Playwright MCP Server, Nicolas Hinternesch
  The author demonstrates several very useful scenarios for automating analytics tasks. I hadn’t previously grasped the power of an LLM + MCP browser automation for monitoring & QA.
• Implementing The Google Tag/GA4, John Charles Tysse
  A conceptual and practical history of the Google Tag. My favorite section heading? “The Google Tag, the Google Tag tag, gtag, Google Tag Manager,” aka, “What happens when engineers get to name stuff”.
  
  

Ideas

• AI’s Holes in the Digital Ozone, Beth Woodcock
  If you’ve read my last couple of roundups, you know I’m a bit giddy about the power of AI agents. This article explores the dark side of that power. 
• The Reflexive Supply Chain: Sensing, Thinking, Acting, Sagar Paul
  The setup of this article is a postmortem on supply chain failures during the pandemic. The author makes the argument that better integration between data inputs and business decision making could have prevented supply chain breakdowns and resulting shortages. That oversimplification almost lost me, but once I got to the section on ‘A Thousand Brains…’, I was hooked. You need to make it through the setup to follow along, but ultimately this is a case for how the flow of data and AI can (and should) support distributed decision making. I like that vision way better than executives replacing knowledge workers with an army of AI agents.
• Agents don’t click: Rethinking Digital Analytics, Patrick Hegnauer
  I touched on this topic in my last roundup, but Patrick goes way deeper. He got me thinking about how we are in the wild west of AI agents. A few sites block them entirely, but mostly we just let them have their way with our content and functionality. It stands to reason that this will change. Matt Prince of Cloudflare recently shared that OpenAI crawls 1,500 website pages for every website visitor it sends, and this ratio has been getting worse fast. I was a bit snarky about Cloudflare’s ‘pay per crawl’ feature above, but they are absolutely on the right track in helping publishers control access.
  Hard to say where it will end up, but I enjoyed Patrick’s thought experiment on the future of measurement.

Miscellaneous

• The EU AI Act Explained, Juliana Jackson
  Speaking of the risks of AI…
  This wasn’t really on my radar, but has major implications for how AI is used in organizations – the EU AI Act prohibits some uses of AI altogether, and regulates others proportional to risk level. If you have or plan to have any business interests or customers in the EU, this is a must read.