Start free
Kunads / Glossary / Learning phase (Meta)

Learning phase (Meta)

Updated 2026-11-19 · Editorial policy

The learning phase is the period when Meta's delivery system is still exploring how best to deliver an ad set, either after it is created or after a significant edit. Meta says ad sets usually exit after about 50 results in the week after the last significant edit. Until then, results swing more and cost per result is usually higher.

How long does the learning phase last?

As long as it takes the ad set to collect about 50 results after its last significant edit, which is a matter of budget and cost per result rather than a fixed number of days. Meta ties the exit to that volume of results within a week. The arithmetic follows directly:

Weekly budget needed to exit ≈ 50 × expected cost per result

Daily budget ≈ (50 × expected cost per result) ÷ 7

Meta does not publish this formula; it is the plain arithmetic of its 50-results guidance.

Worked example: will this ad set leave learning?

A shop expects a cost per purchase of about €30 on a new ad set optimised for purchases.

Options Meta lists for that situation: combine ad sets, widen the audience, raise the budget, raise the bid or cost control, or optimise for a more frequent event such as add to cart. At €100 a day, merging two similar ad sets of €100 each into one of €200 gets close: €1,400 ÷ €30 ≈ 47 purchases a week.

Bar chart: 50 purchases a week are needed to exit learning; one ad set at €100 a day buys about 23, two merged into €200 a day buy about 47.

The same rule for a lead-gen account

Cheaper results make the threshold easier. A home-services company optimises for lead form submissions at about €12 per lead. 50 × €12 = €600 a week, or about €86 a day. Its €100-a-day ad set should leave learning within the week. If the same company optimises for a booked appointment instead, at around €60 each, it needs 50 × €60 = €3,000 a week, about €429 a day, to exit. The deeper event may be the better signal of value, but on a €100 budget it brings only €700 ÷ €60 ≈ 12 results a week, so the ad set stays unstable. Here the trade-off is the choice of optimisation event, not the budget.

What counts as a significant edit?

Any change to targeting, creative or optimisation event, adding a new ad, changing the bid strategy, or pausing for 7 days or longer. That is Meta's own list:

Google's version is shorter and narrower

Google Ads Smart Bidding shows a "Learning" bid strategy status after major structural changes: a new or reactivated strategy, campaigns, ad groups or keywords added to or removed from it, or a significant change to conversion goals. Google says the status never lasts more than 7 days, and that changing a target (for example a target ROAS from 500% to 400%) never triggers it. That differs from Meta, where a large change to a cost or ROAS goal can count as significant. Google adds that performance may take one to two conversion cycles to settle. So the same habit, "don't touch it for a week", means different things on the two platforms: on Meta it is about result volume, on Google about structural changes.

Why it matters for spending

Habits that keep ad sets in learning

In Kunads, budget rules move in 20% steps with one change per campaign per 7 days (a raise after a raise may come after 3 days), which keeps edits small and spaced. For timing in more detail, see how long the Meta learning phase lasts.

Related terms

Changes wait for you. Kunads proposes budget moves in 20% steps with the reasoning attached; nothing reaches Meta or Google until you press Approve.

See what it would change Begin a free trial

Stuck in learning? Short answers

What does Learning Limited mean?

Meta shows it when an ad set is unlikely to get about 50 optimisation events in the week after its last significant edit. Meta names the usual causes as a small audience, low budget, low bid or cost control, or a rare optimisation event.

Do Shops ads follow the same 50-result rule?

No. For Shops ads, Meta says the learning phase completes after at least 17 purchases through your website and 5 through Meta.