What Meta ads actually cost to run, and the floor below which they cannot work

Meta's delivery system needs roughly 50 conversions per ad set per week before it stops guessing. Below that it is working from too small a sample to tell a good impression from a lucky one, so costs stay unstable and results read as noise. That single threshold is where every budget question on this page starts.

The weekly version is simpler to hold in your head: target CPA multiplied by 50. At a €15 cost per conversion that is around €750 a week before the system has enough to work with. Divide by seven for a daily figure, so a €20 CPA needs about €143 a day and a €200 CPA needs about €1,430 a day, for one ad set. Nothing about the second figure is unusual or unfair; it is the same formula with a different input.

Two consequences follow, and both catch people out. The first is that a budget which works comfortably for a cheap product fails completely for an expensive service, because their conversion costs differ by an order of magnitude, so advice borrowed from a different price point is worthless. The second is that the floor applies per ad set, not per campaign, so splitting a budget across four ad sets means dividing your signal by four and clearing the threshold in none of them.

The other number that matters is what you can afford, which is gross margin per sale rather than the sale price. Take a €100 order as an example: at a 30 per cent margin it gives you €30 to spend acquiring it, not €100. Use your own margin rather than that illustration. When the affordable figure sits below the floor the campaign cannot pay for itself at any budget, and that is a pricing or product conclusion rather than an advertising one.

Minimum budget is defined by how many conversion signals you need per week for stable optimisation, so it depends on your cost per conversion and on how deep in the funnel you are optimising. There is no flat figure, which is why the €5-a-day advice circulating online produces accounts that never leave learning.

Work it out in this order:

1. Estimate your expected cost per conversion. If you have never advertised, use your break-even figure as the ceiling and assume the real number will be worse at first. The break-even calculator gives you both.
2. Multiply by 50 for the weekly signal the algorithm wants, then divide by seven for a daily figure. That is your floor for a single ad set.
3. Compare it with what you are willing to spend. If the floor is higher, you are choosing between spending more, optimising for something cheaper, or waiting.

Treat the result as a floor rather than a recommendation. A budget at the floor buys you one ad set that can learn. It does not buy creative testing, audience comparison, or a retargeting layer, all of which need their own signal.

When budget is below the learning threshold and cannot move, you have two levers: reduce what you are asking the algorithm to do, or change what you are optimising for. Neither fully replaces more budget, but both shift the constraint, and doing nothing leaves you paying for a campaign that cannot work.

Consolidate to one ad set. Split budgets multiply the problem, because each ad set needs its 50 conversions a week independently. One ad set at your full budget is always better than two at half. Meta's own testing puts a broad ad set at around 16 per cent better cost per acquisition than fragmented interest ad sets, so consolidation is not just a workaround for a small budget. It costs nothing to do, and it is the fix most likely to feel wrong, because running a single ad set reads as giving up control.

Move up the funnel. If Purchase is not reaching 50 a week, optimise for Add to Cart or Initiate Checkout instead. Those fire three to five times more often, which gives the system enough signal to stabilise. You are trading signal precision for signal volume, which is the right trade when volume is the binding constraint.

Raise the target temporarily. If your break-even CPA is €40 and you are targeting €25, the arithmetic does not close at any budget. Test at break-even for two weeks before judging whether the channel works.

Accept a longer evaluation window. Below threshold, week-on-week comparisons are noise. Judge over three to four weeks rather than seven days, and resist editing in between, because every edit restarts the clock.

Cut weak creative faster. At low volume every impression counts for proportionally more, so one underperforming ad drags the ad set down harder than it would at scale.

Budget does not buy performance directly. It buys the number of things you can learn at once, and that is a more useful way to plan.

• At the floor for one ad set: one audience, one offer, a handful of creatives competing inside that ad set. You will learn whether the offer works. You will not learn why.
• At roughly twice the floor: a second ad set becomes defensible, so you can run a genuine test of audience or offer rather than guessing between them.
• Well above the floor: prospecting and retargeting can be separated, creative testing gets its own budget, and you can afford to keep a proven ad running while testing alongside it.

Scaling past that point has its own failure mode, which is covered in scaling without breaking performance. Getting the structure right at the floor is covered in how many ad sets to run.

Can I run Meta ads on 5 euros a day?

You can spend it, and you will get impressions and clicks. Whether the algorithm can optimise depends entirely on what a conversion costs you. If your conversions cost a euro or two, €5 a day is genuinely workable. If they cost €50, you are buying delivery without ever reaching the volume that lets the system improve, and the money produces data you cannot act on.

Does the 50 conversions a week apply to the campaign or the ad set?

The ad set. This is the most expensive misreading of the rule, because it makes splitting budgets look free. Four ad sets sharing a budget that would clear the threshold once will clear it zero times, and the account performs worse than if you had run one.

Should I optimise for a cheaper event to get more signal?

When volume is what is limiting you, yes, temporarily. Add to Cart or Initiate Checkout fire several times more often than Purchase, so the system stabilises on something. The cost is precision: you are telling Meta to find people who add to cart, and some of those never buy. Move back down the funnel once purchase volume supports it.

Is a bigger budget always better?

No, and past a point it actively hurts. Raising spend pushes delivery into less efficient inventory and broader audiences, so cost per result rises even though nothing else changed. There is a band where your account performs best, and finding it is what the scaling guide is about.

How long should I give a campaign before judging it?

Long enough to clear learning, which depends on how fast conversions arrive rather than on a fixed number of days. Below threshold, judge over three to four weeks. Reporting delay adds to that: conversions can be attributed one to three days after the click, so the last few days of any window are incomplete. Attribution windows covers why.