Meta Ads
The Meta learning phase
Every new ad set starts by guessing. How quickly it stops guessing depends mostly on arithmetic you can do before you launch.
When an ad set is new, or has just been changed in a significant way, Meta's delivery system does not yet know who is likely to convert. It explores. Results during this period swing more from day to day and usually cost more. That period is the learning phase.
The number that matters
The widely quoted guidance is that an ad set needs about 50 of its optimisation events within 7 days to leave learning. The event is whatever you told it to optimise for. If that is leads, it needs about 50 leads in a week. If it is purchases, about 50 purchases.
Treat 50 as a guide, not a law. What matters is the principle: the system learns from results, so an ad set that produces very few of them learns slowly.
Work out your budget first
Learning phase budget check
At this budget the ad set is unlikely to leave learning. Consider optimising for an event that happens more often.
Learning limited is not a penalty
If an ad set is not getting enough results, its status changes to Learning limited. Your ads keep running. The label is a forecast that the ad set will not gather enough data at its current pace, and it is common on small budgets.
Edits that restart learning
Significant edits send an ad set back to the start. The usual list:
- Changing the targeting, placements or optimisation event
- Adding a new ad to the ad set
- Changing the bid strategy
- A large budget change, commonly put at more than about 20 percent in one go
- Pausing for a week or longer
The practical rule is to batch your changes. Make them together, then leave the ad set alone long enough to judge it.
If your budget is small
- Consolidate. One ad set with the whole budget learns faster than four sharing it. Splitting spend is the most common reason small accounts stall.
- Optimise for something that happens often enough. If you get ten purchases a week, the system has little to learn from. A lead, or an earlier step, may give it more signal. Move down the funnel as volume grows.
- Keep targeting broad. Narrow audiences cap how many results are possible.
- Be patient on purpose. Decide in advance how long you will wait and how much you will spend before judging a test.
For local service businesses. A contractor spending a modest amount per day will rarely hit 50 leads a week, and that is fine. Run one campaign and one ad set, optimise for leads, change little, and judge on cost per booked job over a month.
Want these numbers working for you?
Tell us what you want to raise, sell or book, and we will show you what we would do first.