A new campaign needs about a week before its numbers mean anything, and on a small budget it needs longer. The first days are a test run for Meta's system. It is trying your ad on different people to see who responds. During that time one day can look great and the next can bring nothing. Judging on day three is the most common way owners talk themselves out of ads that would have worked, or into ads that will not.
What happens in the first days?#
- Review. Meta checks the ad before it runs. It says most ads are reviewed within 24 hours.
- The learning phase. Meta describes this as the period when its delivery system still needs to learn how an ad set may deliver and perform. An ad set is the part of a campaign that holds the audience and the budget.
- Settling. Meta says performance usually stabilises after about 50 results in the week after the ad set's last significant edit. A result is whatever you asked for: a lead, a purchase, a click.
Why is a small budget slower?#
Because 50 results in a week is a lot when each one costs real money.
| Daily budget | Spend in a week | Leads in a week, if a lead costs £20 |
|---|---|---|
| £10 | £70 | 3 or 4 |
| £20 | £140 | 7 |
| £50 | £350 | 17 or 18 |
| £145 | about £1,000 | 50 |
At £10 a day and £20 a lead, you reach three or four leads a week. The system wants about 50. Jon Loomer, who has written about Meta ads for many years, makes the same point with dollars: at $10 a day you would need a result for every $1.40 spent to leave the learning phase.
So on a small budget, most campaigns never fully leave it.
What does "Learning limited" mean?#
It is the label Meta shows when an ad set is unlikely to get about 50 results in a week. Meta says plainly that learning limited is not a penalty. Your ads still run. Meta describes it as a sign that its system cannot make the most of the budget with the current setup.
Meta's fixes include:
- Combine ad sets. One ad set with the whole budget learns faster than three that share it.
- Widen the audience. A larger area or fewer restrictions gives the system more room.
- Raise the budget, if your sums allow it.
- Aim for a result that happens more often. For a shop, Meta's example is aiming at "add to basket" in place of purchases.
For most small businesses the first fix is the useful one. Run one campaign with one ad set and put all the budget there.
What should you not touch in the first week?#
Meta says only a significant edit sends an ad set back into the learning phase. These always count as significant:
- any change to who the ad is shown to,
- any change to the ad's picture or words,
- changing the result you are aiming for,
- adding a new ad,
- pausing for seven days or longer,
- changing the bid strategy, which is the rule for how Meta bids for you.
Budget changes depend on size. Meta's example is that going from $100 to $101 is unlikely to reset the learning, and going from $100 to $1,000 may.
So leave the campaign alone for seven days. If you must change something, change it once and start the week again.
So how long should you wait?#
- One week before you look at the figures with any confidence.
- Three to four weeks on a budget of £10 to £20 a day before you decide whether ads work for you. By then, at £20 a lead, you may have about 10 to 28 leads, which is enough to see a pattern.
- Stop sooner only if something is clearly broken: the ad is not being shown, or people click and the page fails. See Facebook ads not working?.
Some guides tell you to judge after three to five days. Whito's 2026 UK guide says to give it three to five days before judging results. That is enough to spot a broken ad. On a small budget it is too short to judge the cost of a lead.
"It worked for three days, then died"#
This is the learning phase seen from the owner's side. Early on the system shows your ad to the people most likely to respond, and a few do. Then it has to find more, on a budget that buys very few results a day. A quiet spell follows. Three good days and four quiet ones are one average week. Look at the week, not the day.
What Madder does about it#
Madder is an AI agent that sets up and looks after Facebook and Instagram ads for a small business. It is built to be patient with a new campaign.
Its daily check-up, which is off until you switch it on, does not suggest any change to a campaign that first spent less than seven days ago. It marks it as still learning. After that it suggests a change only when the numbers from the last settled week are clear. A pause, for instance, needs at least three settled days with no leads and enough spend to make that meaningful. A budget change is 20% at most. On a quiet day it suggests nothing.
Madder cannot make the learning phase shorter. Nobody can. It is in early access with a waitlist. See Autopilot and the daily check-up and How Madder works. For setting the budget itself, see How much should you spend on Facebook ads?.
Questions people ask#
How long does it take for Facebook ads to start working?
Allow a week before the figures mean much. Meta says performance usually stabilises after about 50 results in a week. On a small budget, allow three to four weeks before deciding.
What is the Facebook ads learning phase?
It is the period when Meta's delivery system is still learning who responds to your ad. Meta says it usually ends after about 50 results in the week after the last significant edit.
Is £10 a day enough to get out of the learning phase?
Usually not. At £10 a day you would need each result to cost about £1.40 to reach 50 in a week. Your ads still run, with less steady results.
Does editing my ad restart the learning phase?
Yes, if the edit is significant. Meta lists changes to the audience, the ad's picture or words and the goal among those. Small budget changes may not restart it.