My roundup of recent analytics news and ideas that I find particularly useful and/or interesting. I mostly work in the Google analytics stack and I’m a little bit obsessed with the intersection of AI and analytics.
It Has a Name, and Its Name is “Data Mesh”
“… a word is nothing but a painting of a fire. A name is the fire itself.” – Patrick Rothfuss
I recently spent some time trying to articulate my philosophy of marketing data management. I started with the idealized vision of a “single source of truth”, and the centralized control it requires. My hot-take on that is: sounds great, but that approach can never keep pace with the evolving landscape of marketing data and needs.
What I generally experience in practice is a chaotic mess of spreadsheets, reports and dashboards, overseen by marketers who are overwhelmed and underqualified to properly manage data.
The right approach, I believe, is that marketing teams have the resources and training to manage their own data, versus relying on a data team that lives within a technology organization. The central data team in this model adopts a “teach-you-how-to-fish” mindset, versus “we-fish-for-you”.
I was part-way through laying out my perspective when I read this description of a data mesh:
“Traditionally, we’ve had central data platforms: one big warehouse, one big lake, one central team that ‘owns all the data’. The idea was to create a single source of truth. Sounds great. Until it doesn’t scale. The central team becomes a bottleneck. Domain experts (like Marketing or Ops) have to file Jira tickets just to get a column renamed. Meanwhile, data quality drops because the people managing the data aren’t the ones generating or using it day-to-day.
Data Mesh changed this — instead of one central team owning everything, each domain team owns its own data — and treats it like a product.”
I couldn’t have said it better myself! I still plan on putting my thoughts into words, but it is pretty exciting to discover that my philosophy has a name. Having a name gives it legitimacy and power. Those will come in handy when I find myself having to convince business leaders that there is a better way.
Product Updates
- Google Ads and Google Analytics: Google’s AI Aadvisors: agentic tools to drive impact and insights
AI-recommendations and actions will be generally available to English-language accounts in December. I’m pretty cynical about the Google Ads Advisor, given Google’s track record of making recommendations that prioritize Ads revenue over customer value. But I’m looking forward to the Analytics Advisor. - Google Analytics: various enhancements to enriching GA from external data sources were announced in November. None of them are significant enough to write about here, but I do think it’s noteworthy that recent updates have less to do with core reporting and analysis features of GA, and more to do with integrating other platforms. Per my intro, I’m not really a fan of the single-source-of-truth mindset, but even if I was I wouldn’t attempt it in GA.
Analytics Workflow
- How to Monitor AI Bots in the Log File Analyser, Screaming Frog
Whether you like it or not, people are interacting with your website content without ever visiting your website. When a person asks an LLM a question, that LLM will often send an agent out to gather information relevant to the prompt. This can range from high-funnel informational queries down to low-funnel price comparisons and the like. That agent visit is filtered from GA4, so right now you probably don’t see it. This article describes how to start reporting on AI bot traffic in Screaming Frog. - From Clicks to Insights: Building Google Analytics User Paths with R and BigQuery, Arben Kqiku (guest posting on Simmer)
A variety of GA path analysis use cases, with sample code. It’s fairly ecommerce-centric, but most of what is described would apply to a site that’s lead-focused too. The analyses are done in R, but are fairly easy to translate into SQL or python if that’s what you prefer.
Data Visualization
- mind the gap: how to represent partial data, Amy Esselman
Various methods for representing time series data inclusive of the current, in-progress month, quarter or year. - Why Amazon charts beat yours, Sven Balnojan
The five principles Amazon execs follow to create data visualizations that drive real business change. The article includes detailed recommendations and provides an interesting window into how analytics help drive Amazon’s success.
Ideas
- Eroding the Edges: (AI-Generated) Build vs. Buy and the Future of Software, Joe Reis
A smart analysis of how AI is changing the build-vs-buy decision, from a person who regards AI coding with a healthy degree of skepticism. - Hyperproductivity: The Next Stage of AI?, Steve Newman
“A hyperproductive individual does not do their job; they delegate that to AI. They spend their time optimizing the AI to do their job better.” The article describes the types of people who are gravitating towards this type of work, and some of the mechanics of being hyperproductive. As the author says, it sounds exhausting, but then working on a large and slow-moving team is also exhausting, in a different way. - 10 Must-Know Concepts Every Analyst Should Know, Olga Berezovsky
I generally try to avoid sharing content that requires a subscription, but I’m making an exception here because: - This is a really great article.
- If you are considering subscribing to anything in the digital analytics world, I recommend this newsletter. She writes from an informed, strategic, and personal perspective. And she’s funny.
Privacy
- The European Commission proposes a new Digital Package to simplify EU digital rules and boost innovation, European Commission
This is a proposed overhaul of GDPR, including new regulations to address the use of personal data in AI more specifically. Reactions I’ve seen range from very negative to neutral. The negative reactions center around weakening personal privacy protections. My personal opinion is that making compliance easier will benefit consumers as well as businesses, so I probably fall in the neutral category, though I don’t yet understand all of the ins and outs. One thing I really like: the proposal outlines a “one-click consent” framework along the lines of GPC (Global Privacy Control).
Miscellaneous
- The AI Search Manual, iPullRank
This is an epic document full of well-researched and detailed information on how LLMs work, how user behavior is changing and a variety of related topics. If your curiosity isn’t satisfied when an SEO consultant tells you “optimizing for AI is pretty much the same as SEO”, this is for you. But notify an emergency contact before diving in – this is a very deep rabbit hole. - Turning Raw Data Into Gold: A Modern Checklist for Data Product Managers, Juice Analytics
Whether or not you think of yourself as a data product manager (I didn’t, but now I kinda do), this checklist is useful for analysts, engineers and consumers of data.