What is Campaign budget optimisation?

Campaign budget optimisation moves the budget up a level. Instead of each ad set carrying its own daily or lifetime figure, one budget sits at the campaign and the platform distributes it across the ad sets beneath according to where it expects the money to work hardest.

Meta's implementation is the one most people mean by the phrase, and Meta now calls it Advantage campaign budget; the older name is what the industry still says. The same pattern exists elsewhere under different names, and the behaviour is consistent: pooled budget, platform-decided split, allocation refreshed continuously rather than set once.

Nothing else changes. Bidding works as it did, targeting works as it did, and the ad sets remain distinct. The only question the setting answers is who decides how much each gets.

Better allocation than a person can manage. The split is recalculated continuously against current conditions. An advertiser reviewing budgets weekly is making the same decision on stale information and far less of it.

Faster escape from learning. Optimisation events pool. Five ad sets each producing a handful of conversions are five places short of the volume a model needs; pooled, they may clear the threshold in aggregate. This is frequently the largest practical benefit and the least discussed.

Less administrative work. One number to manage instead of several, and no weekly ritual of shifting money between ad sets.

Responsiveness. When an audience fatigues or a competitor arrives, the allocation moves without anybody noticing first.

The characteristic complaint is that one ad set takes nearly everything and the others sit idle.

That is the mechanism working. The platform is told to get the best result from the budget, and if it believes one ad set will produce it, sending the money there is the correct execution of the instruction.

Two things make it a problem rather than a feature.

The first is early concentration on thin evidence. The allocation decision is made before much data exists, and once made it is self-reinforcing: the favoured ad set accumulates delivery and therefore data, while the others accumulate neither and never get the chance to prove otherwise.

The second is that the platform cannot see reasons that are not in the objective. An account may be running a second audience to keep it warm for a launch, to maintain a presence in a market it intends to grow, or to gather data for a future campaign. None of that is visible, and the money will leave.

Where the platform offers spend floors and ceilings at ad-set level, they are the sanctioned answer: the pool is preserved, and each ad set is guaranteed a share. They cost something in allocation efficiency and they buy back the ability to keep a deliberate structure alive.

The blunter alternative is separation. Ad sets that must be funded regardless of performance belong in their own campaigns with their own budgets, which restores the split to the advertiser completely and gives up the pooling entirely.

The question that decides between them is whether the structure encodes a decision worth protecting. A split across three lookalike audiences is not a decision; it is a hypothesis, and the platform is better placed to resolve it. A split between prospecting and retargeting, or between two markets with different strategic weight, frequently is a decision, and handing it to an optimiser that cannot see the reasoning will produce an efficient answer to the wrong question.

Testing hypotheses. Several audiences, several placements, several creative angles, where the objective is to find what works and the account has no prior commitment to any of them.

Thin volume scattered across ad sets. Pooling to escape learning is worth a great deal here.

Ongoing performance campaigns where the only objective is the result and no ad set has a strategic role of its own.

Where an ad set has a job the objective does not describe. Keeping an audience warm, maintaining share in a market, or gathering data for later are all invisible to the allocator.

Where audiences overlap heavily. Pooled budget across overlapping audiences concentrates spend on whichever one wins the internal comparison, and the comparison is partly an artefact of the overlap rather than of merit.

Where budgets are contractual. An amount agreed for a specific market or product is not available for reallocation, whatever the performance.

Where the account is diagnosing something. A fixed split is one fewer moving part.

Switching the setting is material, so the campaign enters a learning period and delivers unevenly for a stretch afterwards.

That has a practical consequence for evaluation: the first fortnight is not the test. An account that turns it on, sees unstable delivery and turns it off has paid for two learning periods and learned nothing. If it is worth trying, it is worth leaving alone long enough to produce a readable answer.

The deeper consequence of pooling is that campaign structure stops being an organisational choice and becomes the main instrument an advertiser has.

With budgets at ad-set level, structure is mostly bookkeeping: how the account is arranged for reporting and for manual control. With one budget at campaign level, structure decides what competes with what. Two ad sets in the same campaign are rivals for the same money; the same two in separate campaigns are not.

That turns several ordinary decisions into consequential ones.

What goes in a campaign together is a statement about which things should be traded off against each other. Prospecting and retargeting in one campaign means the platform may fund whichever converts more cheaply, which will reliably be retargeting, and the account will quietly stop acquiring anybody new.

How many ad sets determines how thinly the events are spread while the allocator is deciding. Fewer, larger ad sets reach conclusions faster and with better evidence.

Whether audiences overlap determines whether the comparison is fair. Two heavily overlapping audiences compared inside one campaign are partly competing with themselves, and the winner is not necessarily the better audience.

The rule that falls out of this: put things in the same campaign when you genuinely want the platform to choose between them, and in different campaigns when you do not. It is a simple test and it resolves most structural arguments in an account running pooled budgets.