Meta Ads learning phase: which edits reset it and how to fix learning limited

If you work with Meta Ads, you have surely seen the word “Learning” in the Delivery column. It is normal for a new campaign to go through it. The problem is going back into that phase by accident: every time it happens, results become less stable and cost per result usually rises.

This guide sums up what Meta says about which edits reset learning, how to tell when the last one happened and what to do if the ad set gets stuck in “Learning limited”.

What the learning phase is

It is the period when the delivery system is still working out how to show an ad set to get the result you chose. It starts when you create a new ad or ad set, or when you make a significant edit to an existing one.

According to Meta, an ad set exits the phase once delivery stabilises, which usually happens after about 50 results in the week after the last significant edit. Until then performance is less stable and CPA is usually higher, so the numbers from those days do not predict how it will perform later.

For Shops ads the threshold is different: 17 purchases on the website and 5 through Meta.

What counts as a significant edit

Meta points out that every edit has some effect on delivery, but only significant edits send an ad set back into the learning phase. These always reset it:

• Any change to targeting.
• Any change to ad creative.
• Any change to the optimisation event.
• Adding a new ad to the ad set.
• Pausing the ad set for 7 days or longer: it re-enters learning when you unpause it.
• Changing the bid strategy.

The ones that depend on how much you change

These may or may not reset learning, depending on the size of the change:

• The budget amount.
• The ad set spending limit.
• The bid control, cost per result goal or ROAS goal.

Meta’s example: raising a budget from $100 to $101 is unlikely to reset learning; taking it from $100 to $1,000 very likely will. Meta does not publish an exact percentage. In practice, many advertisers raise budgets by 20% at most and wait a few days between increases: a working rule, not an official figure.

If you use Advantage+ campaign budget

• Meta distributing budget across ad sets does not send them back into learning.
• A significant edit to one ad set does not affect the other ad sets in the same campaign.
• Adding a new ad set does not reset the others either.
• However, changing the campaign budget or bid strategy can send several ad sets back into learning at once.

For ads about social issues, elections or politics, changing the “paid for by” disclaimer also counts as a significant edit.

How to tell when the last significant edit happened

1. In Ads Manager, select the ad set you want to review.
2. Look for the “Last significant edit” column. If you cannot see it, add it from Columns.
3. Check the Results column: if the ad set is not reaching about 50 results a week since that date, it is unlikely to exit learning.

Logging every change with its date, even in a spreadsheet, makes it much easier to understand later why results moved.

What “Learning limited” means and how to fix it

It appears when an ad set is unlikely to collect about 50 optimisation events in the week after its last significant edit. Meta makes clear it is not a penalty: it signals that budget is not being used well because the system lacks enough data to optimise.

The usual causes, according to Meta, are a small audience, a low budget, a low bid or cost control, high auction overlap with your other ad sets, an infrequent optimisation event or too many ads running at once. The fixes it suggests:

• Combine ad sets and campaigns, so results concentrate in fewer places.
• Expand the audience.
• Raise the budget to a level that can reach those 50 events.
• Raise the bid or cost control if you use one.
• Optimise for a more frequent event, such as add to cart instead of purchase. It brings more volume, but check that those events actually turn into sales.

Once the ad set collects enough events since its last significant edit, it moves from “Learning limited” to “Active”.

You do not avoid the learning phase: you manage it. The goal is to enter it rarely and leave it quickly.

How to edit without starting over

1. Batch your changes. If you are going to touch targeting and creative, do it at the same time: learning resets once, not twice.
2. Do not edit in the first week unless something is broken. Results from those days do not predict performance.
3. Think about where to test new creative. Adding an ad to a working ad set sends it back into learning; it is often better to add them all in a single batch.
4. Raise budgets gradually and leave a few days between increases.
5. Avoid pausing working ad sets for 7 days or longer.
6. Use realistic budgets: Meta warns that one that is too small or inflated gives the system the wrong signal.

Structure matters too: with fewer ad sets, each one reaches 50 results sooner. We cover it in detail in how to structure Meta Ads campaigns with Advantage+.

How we handle it at Semcial

We plan each account’s changes in windows, log every significant edit with its date and build the structure so each ad set has enough volume to exit learning. That way decisions are made on stable data, not on first-week numbers.