The Meta Learning Phase: Why It Exists, How to Exit It and the Budget Formula

An ad set needs around 50 conversions in a week to leave the Meta learning phase. Most small businesses in Lebanon and the Gulf launch with a budget that could never produce that number, then edit the ad set every two days because results look unstable, which resets the clock. The ad set never learns, the costs never settle, and the advertiser concludes that Meta ads do not work.

Key takeaways

• The learning phase is the period after an ad set launches or is significantly edited, while Meta works out how to deliver for your optimisation goal.
• You need around 50 conversions per week to exit it. Below that, the ad set stays in learning and performance is unstable.
• Minimum weekly budget equals target CPA multiplied by 50. A $10 target CPA needs at least $500 a week.
• Four things get you out: enough budget, no significant edits, a broad audience, and seven days of patience.
• Daily and lifetime budgets suit different campaigns, and Advantage Campaign Budget moves money between ad sets for you.

What the learning phase is

The learning phase is the period right after an ad set is launched or significantly edited, during which Meta's algorithm works out the best way to deliver your ads to achieve the optimisation goal you chose. Meta is testing who to show the ad to, on which placements and at what times, and it needs completed results to learn from.

During this phase performance may be unstable and costs may be higher. That is expected, not a sign that something is broken. The delivery column shows "Learning" while it runs and "Learning limited" when results are too few to finish.

Learning happens per ad set, not per campaign, because the ad set holds the audience, placements and budget the algorithm is learning about. Five ad sets means five learning phases, each needing its own 50 conversions, which is why we tell beginners to launch with fewer ad sets and why audience overlap is so damaging: split results mean none of them exits.

The 50 conversions per week rule

You need around 50 conversions per week for an ad set to exit the learning phase, and "conversion" means whatever your optimisation goal counts. If you optimise for leads, that is 50 leads. If you optimise for purchases, 50 purchases. If you optimise for link clicks, 50 clicks, which is easy, and that is partly why Traffic campaigns look so stable while Sales campaigns look so erratic.

The implication: the harder the action you optimise for, the more budget you need to learn. A $30 lead needs ten times the budget of a $3 landing page view to reach the same 50 results.

The budget formula, with two worked examples

Minimum weekly budget equals target CPA multiplied by 50. That is the formula we teach in the program, and it turns the learning phase from a mystery into arithmetic.

Example 1: a skincare brand in Dubai running a Sales campaign. The target cost per purchase is $10. Ten dollars multiplied by 50 is $500, so the weekly budget should be at least $500, or roughly $71 a day. At $20 a day, the ad set would need more than three weeks to see 50 purchases and would never exit learning, because the count is per week.

Example 2: a dental clinic in Beirut running a Leads campaign. The target cost per lead is $6. Six dollars multiplied by 50 is $300 a week, or about $43 a day. If the clinic can only afford $15 a day, it has two honest options: accept that the ad set will stay in learning and judge it on cost per lead anyway, or change the optimisation goal to something cheaper and more frequent, such as landing page views, and accept a less precise signal.

Run the formula backwards and it is just as useful: with a fixed weekly budget of $200, the maximum target CPA you can learn against is $4. If your real cost per result is higher, learning limited is the expected outcome.

The four exit tactics

Four things get an ad set out of the learning phase.

1. Budget. Set the budget to generate 50 conversions per week using the formula above. If you cannot, consolidate ad sets so the budget you do have is concentrated on one.
2. Edits. Avoid significant edits during learning. Every meaningful change restarts the phase.
3. Audience. Use broader audiences initially. A wide audience gives Meta more people to test, so it finds 50 results faster. Narrow later, once the ad set is stable.
4. Timeline. Allow seven days before major changes. Judging an ad set on day two is judging the learning phase, not the ad.

Daily vs lifetime budget, and Advantage Campaign Budget

Daily budget is right for ongoing campaigns and lifetime budget is right for campaigns with a fixed end date, and the choice changes what controls you get.

	Daily budget	Lifetime budget
What you set	An average amount spent per day	A total spend over the campaign duration
Best for	Ongoing campaigns with better daily control	Fixed-date campaigns with flexible daily delivery
Ad scheduling by day and hour	Not available	Available

Ad scheduling by day and hour is only available with a lifetime budget, so a restaurant that wants ads only from 5 pm to 10 pm needs one.

Advantage Campaign Budget (ACB) sets the budget at the campaign level and automatically distributes it across ad sets to optimise delivery. You give Meta the total and it shifts money to the ad sets producing results, a sensible default for a first campaign. Whichever you choose, budget control sits at the campaign or ad set level, never at the ad level.

What not to edit while the ad set is learning

Any significant edit to an ad set restarts the learning phase, so during the first seven days leave the parts the algorithm is learning about alone. In practice that means the targeting, the placements, the optimisation goal, the conversion event, the creative in the ad set, and the budget, which should not be changed by large amounts while learning runs. This is where the 20 percent rule from our optimisation checklist comes from: scale winners gradually with 20 percent budget increases, so growth does not throw a stable ad set back into learning.

What you can do freely: read the reports, use breakdowns to see which placements and age groups are performing, and prepare the next creative so it is ready when the current one fatigues. If you want the diagnostics for that stage, Meta Ads Not Performing? Diagnose It by Frequency, CTR, CPA and Reach walks through them.

Where this is taught

The learning phase, budgeting options and the CPA formula are part of week 11 of the Professional Digital Marketing Associate Diploma, where every student calculates the minimum weekly budget needed to exit learning at their own target CPA before launching their project campaign.

Frequently asked questions

What does "learning limited" mean? It means the ad set is not on track to reach around 50 conversions in a week, so Meta cannot finish learning. The fixes are the four exit tactics: more budget, a broader audience, fewer edits and more time, or a cheaper optimisation goal.

Does the learning phase apply to the campaign or the ad set? The ad set. Each ad set learns separately, which is why splitting a small budget across many ad sets keeps all of them in learning.

Can I run ads with a budget below the formula? Yes, and many small businesses do. The ad set may never exit learning and costs will be less stable, but a campaign judged honestly on cost per lead can still be profitable. The formula tells you what stable delivery costs, not whether you are allowed to advertise.

Does changing the ad copy restart learning? Editing the creative inside an ad set is a significant change and can restart learning. The safer route is to add a new ad alongside the existing one, or to test creative in a separate ad set.