Unmodeled LAB #PilotResults being written upFree
Build Your Own AI Growth Analyst
· 10:00–14:30 · SB Hotels Plaza Europa
One morning in Barcelona. 20 marketers try to build an AI growth analyst on real marketing data, in the room, together. I document what happens and publish it as research: what got built, what broke and what worked.
The question
Can 20 marketers build a functional AI Growth Analyst in one morning?
What you receive
- The prompts used during the training
- The n8n automations used during the training
- All other assets used during the training
Is it for you?
Requests are reviewed by hand. This is who I have in mind:
- You work in marketing, growth or ecommerce and want to build with AI, not hear about it.
- You’d rather test an idea in front of peers than watch a demo.
- You’re fine with the session being documented and published as research.
Everyone in the room takes part: you bring the marketing questions you deal with, build alongside other marketers and compare what works.
How the morning works
Proposed format
Inspired by the spirit of MeasureCamp: the people in the room bring the questions and do the work. Here it all turns around one question. The exact agenda is still being closed.
- 01
Frame
We start from the question and agree what “functional” has to mean by the end of the morning.
- 02
Build
People work in small groups on real marketing data. The room decides what to try.
- 03
Show
Each group shows what it built and where it got stuck. Peers question it.
- 04
Document
I capture what was built, what broke and what worked, for the write-up.
What I’ll look at
Proposal, not results
These are the criteria I plan to use when writing up the morning. If they change, the write-up will say so. The verdict isn’t decided in advance.
- Does it work end to end?
- From a growth question to an answer, without someone fixing it by hand on the way.
- Is the answer right?
- Checked against the data it was built on, not against how convincing it sounds.
- What did it take?
- Time, steps and the points where people got stuck.
- Where did it break?
- Failures are part of the result, not something to leave out.
Example only
“Which channel brought last month’s new customers, and did that change from the month before?”
An analyst built that morning would pass “Is the answer right?” only if its answer matches what the raw data shows.
To show what the criteria mean. It isn’t the agreed scope or a result.
One morning with one group is not a controlled trial. The write-up separates what was measured, what was observed and how I read it, as every piece does. How I label evidence →
Results and resources
The LAB has happened and I’m writing it up. The write-up will appear here and in the newsletter.
Not in the room?
Get what happened.
The LAB write-up goes to newsletter subscribers first.