The learning phase is the period after you create a Meta ad set or make a significant edit, while Meta's delivery system works out how to deliver it. It usually ends after about 50 results in the week after the last significant edit, and results are less stable until then.
FormulaWeekly budget to exit learning ≈ 50 × cost per result
The learning phase is the period after you create a Meta ad set or make a significant edit, while Meta’s delivery system works out how to deliver it. It usually ends after about 50 results in the week after the last significant edit, and results are less stable until then.
How long does the learning phase last?
Until the ad set delivers stably, which Meta says usually happens “after about 50 results in the week after the ad set’s last significant edit”. The Delivery column reads Learning meanwhile, per Meta’s About the learning phase page.
“Results” means your optimization event, so a campaign optimizing for purchases needs about 50 purchases in 7 days. Shops ads have their own rule: 17 purchases through your website and 5 through Meta. During learning, costs per result are usually higher and less stable, so early numbers aren’t a forecast. Add the Last significant edit and Results columns in Ads Manager to see where an ad set stands.
What does 50 results a week cost a real store?
More than most small stores spend. Using real median item costs from public Shopify catalogs, read on 18 October 2026, and an assumed 50% margin, 50 purchases at break-even cost run from about $750 to $3,475 a week.
A Glossier-like store selling $30 items breaks even at $15 a purchase, so 50 purchases is $750 a week. A Rothy’s-like store at $139 needs about $3,475. Below those budgets, an ad set optimizing for purchases is likely to stay Learning limited. That’s why the cost per acquisition you can afford, not just your daily budget, decides whether learning can finish. Combining ad sets under one campaign budget pools those results.
Which edits restart learning?
Any change to targeting, creative or optimization event, adding an ad, a pause of 7 days or longer, and a new bid strategy. Budget, bid and spend limit changes restart it only if they’re big, according to Meta’s significant edits page.
Meta’s example: a budget move from $100 to $101 is unlikely to matter, while $100 to $1000 may reset one or more ad sets. So batch your edits, scale budgets in steps, and add new creative as a new ad set rather than editing a learning one. Meta says you shouldn’t try to avoid learning completely; testing is how results improve.
What does Learning limited mean?
It means the ad set is unlikely to reach about 50 optimization events in a week, so delivery can’t optimize. Meta’s Learning limited page says it isn’t a penalty.
Meta lists 5 fixes: combine ad sets and campaigns, expand the audience, raise the budget, raise the bid or cost control, or choose a more frequent optimization event, for example moving from purchases to add to cart. For a small store the last one is often the only affordable route, but check that add-to-cart signals are clean first; the Conversions API helps there. Our Meta ads management team sizes campaigns so learning can finish, and the Meta ads for e-commerce guide covers the rest.