Facebook Ads Budget Calculator

Setting it out

Budget on paid social is set at campaign or ad set level, and the platform spends it within the day rather than pacing to a monthly cap the way search does.

Required budget = target conversions × CPA, with the added constraint that ad sets need enough conversion volume to exit the learning phase, commonly around 50 conversions a week.

That learning threshold is a real budget floor: an ad set below it never optimises properly regardless of how long it runs.

A concrete case

At a $27 cost per acquisition, an ad set needs roughly $1,350 a week, $193 a day, to reach 50 conversions and exit learning.

Splitting a $200 daily budget across five ad sets gives each $40 a day and ten conversions a week. None exits learning, and all five underperform.

Consolidating the same $200 into one ad set gives 52 conversions a week, exits learning, and typically produces a materially better cost per acquisition than the five fragmented ones combined.

That is why account consolidation has been the dominant trend in paid social structure: the algorithm needs volume in one place more than the advertiser needs granular control.

What the number hides

The learning threshold is per ad set, not per campaign, so campaign budget optimisation helps only if the money concentrates rather than spreading evenly.

Frequent edits also reset learning. An account being optimised daily may never leave the learning phase at all, which makes the optimisation itself the cause of the poor performance.

Where to go from here

Structure to concentrate conversions: fewer ad sets, broader audiences, and enough budget in each to clear the learning threshold.

Then stop editing constantly. Changes to budget, targeting or creative reset learning, and an account edited every day is permanently in its worst-performing state.

Why broad targeting now usually beats narrow

The platform's conversion optimisation has improved to the point where it finds the right people within a broad audience more reliably than manual interest targeting selects them.

Narrow audiences also raise CPM, exhaust faster and provide less data for the algorithm to learn from, three disadvantages at once, in exchange for a control that the system no longer needs.

The practical structure that has emerged is broad targeting with strong creative doing the qualification: the advertisement itself filters who responds, rather than the targeting. That inverts a decade of paid social practice and it is now well established enough that accounts still built on layered interest stacks are usually leaving performance on the table.

It distributes spend toward whichever ad set the algorithm expects to perform best, which frequently means one ad set takes nearly all of it.

That is usually the right outcome and it means the other ad sets never gather enough data to be evaluated, so any test running inside a CBO campaign is not really a test.

Genuine tests therefore need separate campaigns with their own budgets, which is slower and is the only structure that produces a readable result.