The Meta learning phase: why ad sets get stuck
In this article
When you create a new ad set or make a big change to a running one, Meta’s delivery system needs time to work out the best way to deliver it. That period is the learning phase; the Delivery column in Ads Manager shows a “Learning” status (source). This article covers how long learning lasts, which edits send an ad set back into it, and what to do when an ad set gets stuck.
1. When the learning phase ends
According to Meta’s help page, an ad set leaves the learning phase once delivery is stable — usually after about 50 results in the week after its last significant edit (source). For Shops ads, it takes at least 17 website purchases and 5 purchases on Meta (source).
While learning, an ad set is less stable and usually has a higher CPA, so results then aren’t necessarily a guide to later performance (source).
To see whether an ad set has enough results yet, add two columns in Ads Manager: Last significant edit and Results (add them through Columns if you can’t see them) (source).
2. Edits that send an ad set back into learning
According to Meta, significant edits include pausing the ad set or changing the optimisation event, the audience or the creative. Changes to the bid strategy or budget may count as significant too, depending on how big they are (source).
So Meta’s help page advises (source):
- Wait until the ad set has left the learning phase before editing it.
- Only edit when you have reason to believe the change will improve performance.
- Don’t create too many ads and ad sets; combine similar ad sets to combine what they learn.
- Set a budget big enough to get enough results, and avoid changing it often.
The page also says you shouldn’t try to avoid the learning phase altogether, because it’s the step the system needs to optimise your ads (source).
3. Learning limited
If an ad set is unlikely to get about 50 optimisation events in the week after its last significant edit, the Delivery column shows Learning limited. According to Meta, this isn’t a penalty, but a sign that the budget isn’t being spent effectively with the current set-up (source).
Common causes, according to the page (source):
- a small audience;
- a low budget;
- a low bid or cost control;
- high auction overlap;
- an infrequent optimisation event;
- running too many ads at once.
4. Getting an ad set unstuck
What Meta’s help page suggests (source):
| Fix | Why |
|---|---|
| Combine ad sets and campaigns | Results come in faster and stabilise sooner |
| Widen the audience | A bigger audience means more people who can complete the optimisation event |
| Raise the budget | Too low a budget makes it hard to get enough optimisation events |
| Raise the bid or cost control | Too low a control makes it hard to get enough optimisation events |
| Switch to a more frequent optimisation event | For example, from purchase to add to basket |
To track it, add the Optimization events column: Columns ⇒ Customize columns ⇒ Optimization events ⇒ Apply. Once the ad set gets enough events, it moves to Active (source).
This article’s note: each fix in the table (budget, audience, optimisation event) is itself an edit that may count as significant (section 2), so bundle changes into one go rather than making them piecemeal. Combining ad sets or setting the budget at campaign level: see CBO vs ABO: campaign or ad set budgets on Meta?.
What AdPilot does here
AdPilot for Meta Ads’ recommendations include ad sets stuck in learning, with a reason; the thresholds are AdPilot’s, not Meta’s (source). Editing budget, audience and placements are among the things AdPilot can propose, and they go through review (source).
Further reading
- Budget not spending evenly: see Facebook ads not spending, or spending too much: why and what to check.
- Choosing a campaign objective: see Meta campaign objectives: six options and when to use them.
Checklist
- Added the Last significant edit, Results and Optimization events columns
- Not editing an ad set while it’s “Learning”, unless there’s a clear reason
- Bundled big changes into one go
- “Learning limited”: checked audience size, budget, bid or cost control, number of ads, optimisation event
- Similar ad sets: considered combining them
💬 Need help? Join the ALFA Software support group on Telegram (ask in your own language) — or message our support bot privately about your account or invoices.
Checked against Meta’s Business Help Center and ALFA’s AdPilot for Meta Ads page on 8 Nov 2026.
Sources
- Meta — About the learning phase: https://www.facebook.com/business/help/112167992830700
- Meta — Last significant edit: https://www.facebook.com/business/help/942374239243867
- Meta — About learning limited: https://www.facebook.com/business/help/269269737396981
- ALFA Software — AdPilot for Meta Ads: https://alfasoftwares.com/en/adpilot-meta