A budget holds three different kinds of money
Most budget arguments are really arguments about which of these three someone is talking about.
Keeping media and production separate is what stops a campaign from looking like it is overspending in its first week every single time, since production lands as one charge before any media has had a chance to run.
Money is spent at a rate, not in a lump
A platform does not spend a total. It spends a rate, and the total is what the rate adds up to by the end of the window. That rate is the number every later check compares against. Divide the media budget by the length of the flight and you get the expected daily rate. Pacing measures actual spend against that rate while the campaign runs, which is the only way an 8% weekly shortfall gets caught before it becomes a 30% underspend that nobody can do anything about. Always-on budgets work the same way with a monthly rate and a review date instead of a flight and an end date.Platforms front-load. A campaign usually spends above its rate in the first days while the platform learns, then settles. That early spike is normal behavior rather than a budget going wrong, and treating it as an emergency is the most common self-inflicted pacing wound.
Two places a budget number can come from
Both starting points are used in practice, and they produce different kinds of budget. From the outcome backwards. You start with the result the campaign needs and multiply by what a result has historically cost you. This gives you a number whose reasoning is visible, so it can be argued with at planning time and judged once the campaign is over. It requires history, or an honest admission that you do not have any. From the available money forwards. You start with what has been allocated and work out what it can buy. This is often the real constraint, especially where the budget is set annually by someone who is not running the campaign. The risk is that the resulting spend has no target attached, so at the end there is nothing to compare the outcome against. The two are not mutually exclusive. A number that arrives from the top can still have a target derived for it, and that derived target is what makes it possible to say later whether the money did anything. The budget planning process works from the outcome backwards and treats a missing history as a reason to run a smaller pilot with a read date rather than a full campaign.What allocation actually decides
Allocation is not one decision. A budget gets split along several axes at once, and each split is a bet about where the money will do the most.
A budget can look balanced on one axis and lopsided on another. An even channel split that puts every dollar against a single profile in a single market is still a concentrated bet.
Ways to split a budget, and what each one assumes
None of these is correct in general. Each one is a reasonable answer to a different situation, and each carries a cost worth knowing before you pick it.
Most real budgets combine several of these. What makes the combination defensible is that each line can say which reasoning it came from.
Every line carries a reason
“Roughly 45%” describes the split without explaining it. A percentage leaves nothing to disagree with, so nobody can revisit it or correct it later. A reason gives them something to hold. Compare the two:
Six months later the reason is what tells the next person whether the split still holds. The percentage tells them only what was decided, not why.
A budget too small to read teaches you nothing
Money buys two things: results, and information about whether the approach works. The second one has a floor. A budget that produces a handful of conversions will not support a target bid strategy, and the result at the end will not separate a real effect from noise. This is worth checking before the split rather than after, because it changes what the split can look like. A budget spread thinly across six channels can leave every one of them below the point where the numbers mean anything, where the same money on two channels would have produced two readable answers. Each platform brief carries its own volume thresholds: Google, Meta, TikTok, and LinkedIn.A reserve is only useful with a release rule
Holding money back is a way of buying the option to react to something you cannot predict at planning time. A reserve with no stated release condition does not do that. It gets spent early, on the first thing that looks urgent, or it sits untouched until the flight ends and quietly becomes an underspend. What turns it into a real option is naming two things in advance: the condition that releases it, and the person who decides that the condition has been met.What makes a budget hard to judge later
Every one of these is a decision made at planning time whose cost arrives at the retrospective.What a budget is not
- A forecast. The budget is what you commit. The forecast is what current pace says you will actually spend, and the two diverge from week one.
- A performance measure. Spending the full budget is not a result. Spending less than planned is not a saving until you know whether it was a deliberate decision or a delivery failure.
- A ceiling on outcomes. More money on a campaign that cannot deliver buys nothing. Underspend is usually a delivery problem before it is a budget problem.
- Fixed once approved. Budgets move between channels while a campaign runs. What matters is that each move is recorded with its date, size, and reason, since an unrecorded shift shows up in the next report looking like a performance change.
Related resources
- Budget planning process The steps that turn these choices into a plan with a rate and a target.
- Budget pacing process Tracking spend against that plan once the campaign is live.
- Budget reporting process Reporting spend, pace, and efficiency on a cadence.
- Marketing campaign brief Where the budget and its line items are recorded.
- Pilot brief For a learning budget with a ceiling, a read date, and a kill or scale rule.
- Approval process Who signs a budget off.
- Glossary: pacing The short definition.