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.
How to Cook a Measurement Plan in 10 Minutes or Less
I moderate the #analytics channel in the SEO Community Slack*, and someone recently asked about best practices for coordinating between paid ads, analytics, and development teams. My response was:
“…a good place to start would be some sort of measurement plan, so it doesn’t come down to opinion-vs-opinion when tasks are being defined and prioritized. It doesn’t need to be overly complex – just an agreed-upon overview of what matters and how it’s tracked.”
In a classic, eat-your-own-dog-food moment, it later occurred to me that I was facing exactly the same challenge with a client and that I should probably take my own advice. But I have a slightly embarrassing confession to make: I’ve never actually created a proper measurement plan.
So I turned to AI for help, and my tools included ChatGPT, Claude, the Google Analytics MCP Server (which I love), and the Stape Google Tag Manager MCP Server. I wrote up my approach, with specific prompt details in this post:
Creating a Measurement Plan With the Google Analytics and Stape GTM MCP Servers
* The SEO Community is amazing, btw. If you have even a little bit of involvement with SEO/AIO, you should join.
Product Updates
- No Google Analytics releases this month (as of August 27). Some pretty minor stuff from Looker Studio, but I do notice a pattern. Since they released modern charts, they’ve been making a lot of little improvements to charting capabilities. I still find myself going to python when I want a lot of control over chart presentation, but maybe they’ll get there….
- BigQuery has been quietly adding a bunch of transfer services. I noticed this when reading their announcement of the upcoming pricing model for 3rd-party services (TL;DR: very reasonable pricing model compared to other options for getting data). I swear Facebook and Salesforce weren’t there the last time I looked. If this keeps going, it could effectively eliminate the need for ETL tools like Fivetran and Supermetrics. Interesting.
Workflow
- Adding Context to Analytics Data Using Google Analytics MCP Server, Nicolas Hinternesch
Per above, I’ve been finding the Google Analytics MCP Server immensely helpful. This article gave me some great ideas for things to try next. In the article, Hinternesch demonstrates three powerful use cases for combining GA traffic data with external sources. - 2025 AEO Tool Comparison: Transparent Ratings for AI Search Visibility, Aimee Jurenka
This is a good rundown of some of the top options for measuring brand visibility in AI platforms.
On a related note, one of the top rated tools, Profound, was used to conduct this fascinating study: The Leading Brands & Domains in AI Search Across 10 Business Categories. The thing I find most intriguing about the study is the variation in results between each platform. We’ve all been so focused on reverse-engineering Google for so long, it’s kind of nice to think that we will need to open our eyes and minds to a broader ecosystem. - Data Governance Tailored for Small Companies, Charlotte Ledoux
Excellent guidance for implementing data governance on a budget. Ledoux provides a pragmatic framework focusing on business-critical elements: defining key data domains, establishing lightweight policies, and assigning simple roles. I’ve actually already used some of her recommendations to improve the measurement plan I described above.
In general, Ledoux’ substack is a treasure trove of data governance knowledge.
Ideas
- OpenAI, Anthropic committed ‘chart crime.’ Is it just sloppiness?, Washington Post
A fun collection of bad charts generated by AI. Note that some of the people interviewed for the article didn’t think they were fun at all. Whether you find them amusing or criminal, it’s a good argument for why we urgently need to be teaching critical thinking skills in schools and workplaces. You need to have/create an account to read the article, but you don’t have to pay for a subscription. - Questions as Competitive Advantage, Helge Tennø
We’ve all heard and/or given the advice that analysis should start with a well-defined question, but not all questions are created equal. This article delves into how to frame questions that will produce meaningful results. The author walks through a detailed, real-world example that got me rethinking some of my recent excursions into customer journeys.
Privacy
- With iOS 26, Safari will counter one of the web’s most invasive tracking methods, 9to5Mac
Safari will enable Advanced Fingerprinting Protection by default across all browsing sessions, significantly expanding privacy protections beyond private mode. Their system actively confuses tracking scripts by muddling the signals used for fingerprinting—my read on this is that the impact may be worse than just making a person untrackable, it may make them appear as multiple people, skewing a variety of user- and session-level statistics.
If you’re wondering whether Google Analytics uses fingerprinting, the answer is no, but many ad platforms do. Given Google’s recent turnabout on the technique, one wonders if their own ad platforms are among them.
If you are curious about how fingerprinting works, this website shows you all of the signals your browser is sending that can be used to uniquely identify you. Very cool. - GTM Consent Mode Best Practices: Study of 1,000 Websites, Phil Pearce MeasureMinds analyzed 1,000 top websites’ Google Tag Manager implementations and catalogued a variety of consent-related issues. Their findings constitute a useful checklist of things to watch out for, and include some specific, technical recommendations.
- German Ruling Prohibits Use of Google Tag Manager Without User Consent, Riccardo Gosteli
Speaking of Tag Manager and consent, there was a recent German court ruling that loading GTM without user consent is a violation of GDPR. If we take the ruling at face value, we can’t use GTM to load functional scripts (for example consent banners) and Google Analytics behavioral modeling is rendered useless, among other things. This seems like an extreme interpretation of the law, but there it is.