In brief
When a campaign isn’t doing well, the temptation is to change everything right away. But every major edit restarts Meta’s learning and makes results unstable. Optimizing means changing at the right time and in small steps.
What restarts learning
- A budget changed by more than 20% at once.
- A change of bid strategy or cap.
- A significant change to the audience.
- A change of the optimized event.
- Adding an ad to the ad set, or a long pause (a week or more).
The method
- Wait for the ad set to exit learning (about 50 results within 7 days), or at least 3 to 7 days.
- Diagnose with the metrics (see the previous guide).
- Change only one thing at a time, in this order: first the visual and the message (often the most effective), then the audience, then the placements, and last the budget.
- For a big change, duplicate the ad set rather than editing the original.
- To increase the budget, do it in steps of 10 to 20% every two or three days.
- Write down what you changed, and why.
Pitfalls to avoid
- Stopping a campaign after 24 hours.
- Changing several settings at the same time.
- Doubling the budget all at once.
- Not writing anything down: you no longer know what worked.
Key takeaways
- Patience: let learning finish.
- One change at a time, in small steps.
- Duplicate to test a big change.
Practice in TRIALO
Duplicate a published campaign, change just one setting (the audience or the visual) and compare the simulated results of the two versions.