Learning phase

The learning phase is the period when Meta's delivery system is exploring how to deliver a new or significantly changed ad set. During this time, it tests different people, placements and times to learn where it gets the best results for your optimisation event. Performance is usually less stable while an ad set is learning.

How an ad set exits learning

At the time of writing, Meta's guidance is that an ad set generally needs around 50 optimisation events, such as purchases, within a seven-day period after its last significant edit to exit the learning phase. Ads Manager shows the status in the delivery column.

If an ad set cannot reach that volume, it may show a "Learning limited" status. This means the system is unlikely to gather enough data to optimise fully with the current setup.

What resets learning

Significant edits can send an ad set back into learning, including:

• changing targeting or placements;
• adding new ads or changing creative;
• changing the optimisation event or bid strategy;
• making a large budget change;
• pausing for an extended period and then restarting.

How to work with it

• Consolidate: fewer, larger ad sets can reach the event threshold more easily than many small ones.
• Choose a reachable event: if purchases are too rare, optimising for an earlier event such as add to cart may help, though the system then optimises for that event instead.
• Edit less often: batch changes rather than adjusting daily.
• Scale gradually: increase budget in steps rather than in one large jump.

Being in learning is normal and not a sign of failure. The goal is to avoid getting stuck there.