01 Write down what happened
A result is not a learning yet.
Separate what you know from what you are guessing, in writing.
Say this happened: Hook B got the most paid subscribers. Useful, and not enough. You still need to ask what was different about Hook B, whether it won all the way to paid or only on clicks, whether the angle underneath it also looked strong, whether one creator was much better than the others, and whether it held up after more spend.
3 hooks for the gym-confusion angle
Hook B, “The hardest part isn’t working out…”
click-through and paid subscribers
watch completion
would the same hook work with another creator?
That last line is the one that matters. A good review separates what you learned from what you are still guessing.
How to see this in Airbridge
- Open Reports → Actuals and add Installs, Start Trial, and Subscribe.
- Group by Ad Creative to see Hook A, B, and C as three rows.
- Read left to right. A hook that wins on installs and loses on Subscribe did not win the test Lesson 01 defined.
You'll see:
whether the “winner” held its lead all the way to paid, or only at the top of the funnel.
02 Turn it into a learning
Write the learning so you can use it in a month.
“Maya did well” is an observation. A learning describes the pattern.
“Hook B won.”
“The creator using a problem-first opening brought fewer installs but more paid subscribers.”
“For the gym-confusion angle, problem-first openings brought higher paid conversion than product-first openings. Test the same opening with another creator before treating it as an angle-level learning.”
Now you know what to repeat and what still needs proof.
Three labels, and why they are not Lesson 06’s
Something worked strongly enough that it goes into the next batch as is.
The signal is promising, but you need another variation before calling it a pattern.
The idea or the execution gave you no reason to spend another test slot on it.
Lesson 06’s Scale, Iterate, Watch, and Stop answer “what do I do with this live creative?” These three answer “what do I carry into the next batch?” A Stopped video can still produce a Keep (the angle was right, the delivery wasn’t). A Scaled video can produce a Test again (it won, but only with Maya).
03 Change the next unknown
Keep what worked. Change one thing you still want to learn.
This is the whole lesson in one line.
keep the angle → test new hooks
keep the hook → test another creator
keep the creator → test another angle
don’t make more versions of the same thing → revisit the message or the audience that is bringing the wrong user
Example
Last batch: angle gym-confusion, creator Maya, hooks A / B / C. Result: Hook B wins on paid subscribers.
The wrong next batch: Maya + Hook B + a new angle + a new format + a new CTA. You just threw away the thing you learned.
A better next batch: same angle, same winning opening, creators Maya / Jake / Sarah.
Now the question is: does this hook work because of the message, or because Maya delivered it? That is how a learning compounds.
How to see this in Airbridge
- Open Reports → Actuals and add Start Trial and Subscribe.
- Group by Ad Group (your angle), then add Campaign (your creator). Same angle, one row per creator.
- Add Date to see whether the winner is still winning this week.
You'll see:
whether the angle holds up across everyone who used it, or only worked when one person delivered it. That answer is next week’s batch.
04 Build the next batch
Turn one learning into 3 to 5 tests.
Each test answers one question. An empty cell is a production order.
Build your next batch
Keep what worked. Pick the one thing you still don’t know. The batch writes itself.
01What did you learn?
02What worked?
03What is still unknown?
Your next batch
Pick the unknown to build the batch.