The Meta ads learning phase: what Meta's own pages say

What the learning phase is

Everything below comes from Meta's Business Help Centre as it read on 9 October 2026. The pages carry no publication date, so treat the numbers as current on that day.

Meta's page about the learning phase defines it as the period when the delivery system still needs to learn how an ad set may deliver and perform. It starts when you create a new ad or ad set, or when you make a significant edit to an existing one. While it lasts, the Delivery column in Ads Manager reads Learning.

Meta is plain about the cost. During the learning phase, it says, ad sets are less stable and usually have a higher cost per action, and results do not necessarily indicate future performance. It also says you should not try to avoid the learning phase altogether: the system needs it, and testing new creative is how performance improves.

The number: about 50 results in a week

Meta says ad sets leave the learning phase as soon as they can deliver stably, and that this usually happens after about 50 results in the week after the ad set's last significant edit. Three details in that sentence are often lost when it is repeated.

• It is approximate. Meta qualifies the figure twice and does not present 50 as a hard switch. The stated condition is stable delivery; 50 in a week is when that tends to happen.
• The clock runs from the last significant edit, not from the day the ad set was created.
• The count is of the event you optimise for. The learning phase page speaks of results; the Learning limited page speaks of optimisation events. In practice that means an ad set optimised for purchases is counted on purchases, not on link clicks.

Shops ads are the stated exception. For those, Meta says the learning phase needs a minimum of 17 purchases through your website and 5 through Meta.

To check where an ad set stands, Meta suggests adding the Last significant edit column in Ads Manager and reading it beside the Results column. This matches the figure in our guide to testing ad creative.

Which edits restart learning

Meta's page on significant edits and the learning phase says every edit has some effect on delivery, but only a significant edit sends an ad set back into learning. It sorts edits into the groups below.

One campaign-level note sits beside that list. With Advantage+ campaign budget, Meta says switching the campaign's bid strategy, or adjusting the campaign budget, might send several ad sets in the campaign back into learning at once.

A pause shorter than seven days is not on Meta's list of significant edits. The page does not go further and promise that a short pause leaves delivery untouched.

Budget changes: what Meta says, and what it does not

The most repeated piece of learning phase advice is a fixed limit on budget changes, usually given as a percentage you must stay under. That rule is not on any of the Meta pages listed at the end of this post. Meta gives no percentage threshold for when a budget change becomes significant.

What it gives is one example. Raising a budget from USD 100 to USD 101 is not likely to restart learning; changing it from USD 100 to USD 1,000 may. Everything in between is left to the size of the change, with no line drawn. A percentage rule may be a sensible habit, but it is an advertiser's rule of thumb, not Meta's published guidance.

The budget guidance Meta does publish is this:

• Use a realistic budget. Meta says a very small or an inflated budget gives the system a poor indication of who to optimise for, and that the budget should be large enough to get enough total results.
• Avoid frequent budget changes, which Meta says can send an ad set back into learning.
• Five times the goal, if you set one. Meta's minimum budget guidance says that with the cost per result goal bid strategy, the daily budget should be at least five times the goal: a USD 5 goal needs at least USD 25 a day. This applies to that bid strategy only. It is a minimum to avoid under-delivery, not a formula for leaving the learning phase.

Percentages do appear on Meta's best practices for the cost per result goal, but they concern the goal amount, not the budget: start 10 to 20% above your average cost per result, and move the goal in steps of 5 to 10%. On budget, that page gives no percentage: it says to set enough that delivery is not limited, and to increase it gradually.

What Learning limited means

Meta says an ad set becomes Learning limited when it is unlikely to receive about 50 optimisation events in the week after its last significant edit. The status can appear at any time after you create an ad set or make a significant edit. For Shops ads the test is the same 17 website purchases and 5 through Meta, after seven days.

In Meta's words, the status "isn't a penalty". It indicates that the budget is not being spent effectively because the delivery system cannot optimise with the current setup. Meta lists these general causes:

• A small audience.
• A low budget.
• A low bid or cost control. Meta says a tooltip on the status tells you when this is the cause.
• High auction overlap.
• An optimisation event that does not happen often.
• Other issues, such as running too many ads at the same time.

The status is not permanent. Meta says a Learning limited ad set that receives enough optimisation events since its last significant edit moves to Active.

What Meta recommends, in order of effort

Meta's fixes for Learning limited, and its advice for getting through learning sooner, come to the same short list. The order below is ours, from least to most disruptive; the steps are Meta's.

1. 01

   Stop editing

   Meta recommends waiting until an ad set is out of the learning phase before editing it, and editing only when you have reason to believe the change will improve performance.
2. 02

   Combine ad sets

   Meta's page on combining ad sets says that when too many run at once, each gets fewer results and spends longer in learning. Its example is three ad sets split by closely related interests, which would do better as one with all three. Use breakdowns in reporting if you still want to see each segment.
3. 03

   Run fewer ads

   Meta's guidance on ad volume says that with too many ads live, each delivers less often, fewer leave the learning phase, and more budget is spent before the system can optimise.
4. 04

   Expand the audience

   A larger audience gives more chances for someone to complete the optimisation event.
5. 05

   Raise the budget, or the bid or cost control

   If either is too low to buy enough optimisation events in a week, Meta says the ad set is unlikely to leave learning.
6. 06

   Choose a more frequent optimisation event

   Meta's example is moving from purchases to add to basket.

Most of these steps are themselves significant edits, or may be. That is the trade: one deliberate restart into a setup that can reach 50 events, in place of weeks in a setup that cannot.

Leaving the learning phase does not make an ad good. It means delivery has settled, and a weak ad settles at a weak result. What decides the cost of each result is covered in our guide to how Meta ranks ads.

Sources

• Meta Business Help Centre: About the learning phase, undated, read 9 October 2026
• Meta Business Help Centre: Significant edits and learning phase, undated, read 9 October 2026
• Meta Business Help Centre: About learning limited, undated, read 9 October 2026
• Meta Business Help Centre: Last significant edit, undated, read 9 October 2026
• Meta Business Help Centre: Combine ad sets and campaigns in Meta Ads Manager to reduce audience fragmentation, undated, read 9 October 2026
• Meta Business Help Centre: About managing ad volume, undated, read 9 October 2026
• Meta Business Help Centre: Best practices for minimum budgets, undated, read 9 October 2026
• Meta Business Help Centre: Best practices for cost-per-result goal, undated, read 9 October 2026