Learning phase: what actually resets it and what’s a myth
Natália Vieira · 2026-10-07
Understanding the learning phase in Meta is essential for optimizing your campaigns and avoiding wasted spend.
Edits that reset learning
- Budget increase above 20%: If you increase an ad set’s budget by more than 20%, learning will reset. Keep the increase within that limit to avoid restarting the learning phase.
- Audience changes: Changes that directly impact your target audience, such as targeting or exclusions, also reset learning. If you need to adjust your audience, consider creating a new ad set.
- Placement changes: If you change placements (e.g., Facebook Feed, Instagram Stories), that also causes a reset. Whenever you make this change, be ready for a new learning cycle.
- Creative changes: Swapping the image or video can reset learning. If you need to test new creatives, do it in a controlled way, keeping the other elements constant.
Edits that don’t reset learning
- Ad set name: Changing the name doesn’t impact learning.
- URL params: Changes to URL parameters won’t reset learning.
- Small bid adjustments: Bid adjustments, as long as they stay within a reasonable range, don’t affect the learning phase.
How to get 50 events in 7 days
To reach 50 events in 7 days without spending a lot, consider:
- Refined targeting: Focus on audiences that have already shown interest in similar products.
- A/B tests: Run A/B tests with different creatives and audiences on small budgets, prioritizing the ones that bring the best results.
- Pixel optimization: Make sure the pixel is set up correctly and collecting data. Improve tracking so events get recorded.
Limited learning isn’t the end
If you’re dealing with limited learning, that doesn’t mean your campaign is doomed. Use the CPA decision rule:
- If CPA is higher than expected, consider adjusting the budget or targeting.
- If CPA is within expectations, keep it running and monitor it until the learning cycle is complete.
Duplicate vs. edit
The decision to duplicate or edit an ad set depends on your goal:
- Duplicate: If you’re testing a new creative or a radically different audience, duplicating may be the best option. This lets the original keep progressing through its learning phase without interruptions.
- Edit: If the change is small (e.g., a copy or image tweak), edit the existing ad set. This can save time and resources while keeping the existing learning intact.
What’s been your experience with changes that reset learning? How many edits have you made before getting a satisfactory result?