The Briefing
Your best buyers are now invisible. Research done inside ChatGPT, Perplexity and Google AI Mode leaves no cookie, no UTM, no referral. Zero-click searches have hit 58%. So the channel doing the persuading often gets zero credit in your report.
AI attribution went mainstream. AI-powered analytics is now used by over 56% of marketing teams, and data-driven attribution grew 44% year over year. The tools got smarter. The question of what you feed them did not get easier.
GA4 is guessing more on your behalf. The 2026 updates lean harder on privacy-safe modelling to estimate performance when users decline tracking. Useful. Also worth knowing: some of your numbers are now estimates wearing the costume of facts.
Meta shipped an AI creative suite with brand memory. It learns from your existing ads so new output stays on brand, plus new AI ad experimentation tools. More variants, made faster. Which means your read on which one won matters more than ever.
The EU AI Act disclosure rule went live on 2 August. Realistic AI-generated people, product shots and backgrounds aimed at EU audiences now need labelling. Fines run to 15 million euro or 3% of global turnover. The duty sits with you, not the vendor whose model you used.
The Build
This week was two jobs stacked on each other. Crunching data to find hidden wins. And writing email promo copy.
They turned out to be the same job.
Here is what happened. I had a promo sequence out. I opened the report and did what everyone does: sorted by click rate. There was a clear winner at the top. Great subject line, punchy copy, most clicks. I was ready to write three more like it.
Then I stopped sorting by clicks and started sorting by what happened after the click.
Different email won.
The high-click email pulled a lot of people who opened a tab and left. The quieter email pulled fewer people who actually went on to do the thing I wanted. Same list. Same week. Opposite conclusions depending on which column I sorted by.
Not all clicks are equal. I have said that sentence in strategy sessions for years. This week I got to watch it nearly cost me a campaign.
So I built a prompt that refuses to let me sort by the easy column.
You are my measurement analyst. I will paste raw campaign data. Do NOT rank by clicks, opens or CTR. 1. For every email, list clicks AND the downstream action that actually matters (booked, bought, applied). 2. Calculate value per click, not volume of clicks. 3. Rank by downstream outcome. Show me where that ranking disagrees with the click ranking. 4. Flag any email with high clicks and low downstream action. Tell me what the copy promised that the next step did not deliver. 5. Tell me what this data CANNOT tell me. List the blind spots and the sample sizes too small to trust. Be blunt. I would rather be corrected than flattered.
Step 5 is the one I would fight for. Every tool on earth will hand you an answer. Very few will tell you the answer is thin.
What worked: Forcing the ranking disagreement into the open. When the click ranking and the outcome ranking disagree, that gap is the insight. It also fed straight back into the copy. The high-click email was writing a cheque the next page did not cash. That is a copy problem I could only see through the data.
What didn't work: My first pass asked for "insights" and got a paragraph of confident nonsense built on eleven conversions. Way too small to mean anything. I had to add the blind-spot step to stop it performing certainty it had not earned. Clean inputs matter more than a clever prompt. Garbage in, confident garbage out.
The Play
If you are winging it on data, or you have data and a nagging feeling you are not squeezing it, do this today.
Pick your last campaign. Write down the metric you normally judge it by. Now write down the thing that actually pays you: booked calls, sales, applications. Then rank your assets by both.
Where the two lists disagree, that is your money. Every asset sitting high on clicks and low on outcomes is a promise your funnel is not keeping. Fix the promise or fix the page.
Do not have the second column? That is the finding. You are not short on tactics, you are short on a scoreboard. Set up one tracked outcome this week. One. The one that pays you.
Paste the prompt above with your own raw numbers. Answer its questions honestly, including the ones about sample size.
Try this today. Open your last campaign report, add one column for the outcome that pays you, and re-rank. If the top row changes, you just found the week's win.
Download analytics-measurement-1.md
Growth Partner for founders stuck between $1M and $20M | I find the money you're sitting on and build the machine that collects it | $1B+ in client results.
Growth Partner for founders stuck between $1M and $20M. I find the money you’re sitting on and build the machine that collects it. $1B+ in client results.