The Briefing
Adobe launched Custom Algorithms, which trains your ad bidding on raw analytics data instead of the thin, filtered stuff that leaks through most APIs. It can weigh real signals like video views and sign-ups. If you buy media, this is the machine finally getting fed the full picture.
Google is rolling out a Generative AI report inside Search Console. For the first time you can see how your content shows up in AI search results, not just the blue links. If you sell through content, start watching this number now.
LinkedIn shipped AI tools in Campaign Manager: Draft with AI and auto ad variants that spin up new headlines and intros in seconds. More variations means more to test. That only helps if you actually read the results.
Google's AI Max left beta, and its new terms let Google generate your targets and ads for you. Handing the machine the wheel is fine. Driving blind is not. You still need the data to tell you if it is working.
The Build
Last week I installed Founder OS. This week the payoff showed up in a boring place: reporting.
Reporting is now streamlined, with deep research and variability testing built right in. Plain version: it pulls the real numbers, digs into the why, and lets me spin up two versions of something and run them side by side. The winner is picked by data, not by whoever talks loudest in the room. No more guessing.
That last part matters more than it sounds. For 15 years I have watched smart founders lose money to a strong opinion held with total confidence. I am one of them. The report is only useful if it can tell me I am wrong.
So I tested a subject line I was sure about. I had a favorite. Clever, a little cheeky, the kind of line I would defend in a meeting.
I made myself write it as a real hypothesis first, using Neil Patel's split-test method. One variable. A clear guess. A number that decides.
You are my testing engineer.
Here is my test:
Control (A): "The report took 3 hours. Now it takes 6 minutes."
Variation (B): "I let the data pick. My gut lost."
Do 4 things:
1. Write the hypothesis in this format:
"Because we saw [X], we believe [change] for [audience]
will cause [metric] to move. We'll know when [result]."
2. Give each an ICE score (Impact, Confidence, Ease, 1-10).
3. Tell me the sample size I need for 95% confidence.
4. Warn me if I'm about to change more than one variable.
Then I ran it. My clever line lost. The plain, honest one won, and it was not close once the sample got big enough.
What worked: Writing the hypothesis before the test. It forced me to name what I expected and pick one metric. When the data came back against me, there was no argument to have. The number just decided.
What didn't work: My first run, I called it early. Version B jumped ahead on day one and I almost shipped it. The sample was tiny. I waited for real significance, and the gap held, but it taught me the lesson again: an early lead is not a winner.
The Play
If you run your business on feel, this one is for you. Instinct built your company. It will not scale it. You are not winning by deciding you know better than the data, you are just paying to find out later. The founders who pull ahead now are the ones who let the numbers overrule them fast.
Pick one thing you have an opinion about this week: a subject line, a headline, a price. Then stop arguing it in your head and test it. Use this to set it up right:
Help me run a clean A/B test on [the thing]. 1. Turn my opinion into a one-variable hypothesis. 2. Name the single metric that settles it. 3. Tell me the sample size for 95% confidence. 4. Stop me if I try to change two things at once.
Try this today. Take the one decision you feel most sure about, write it as a hypothesis, and set the number that would prove you wrong.
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.