Budget allocator template

A budget allocator is one row per destination and one argument per row. A destination is wherever the money can go — a channel, a phase, a segment, a market — and the row records what its share is, what the share is argued from, the least and the most it can usefully hold, and what has to be true before money moves.

That is the whole shape. Everything else on this page is about what belongs in each column, and about why the two columns most allocations leave out, the floor and the reallocation rule, are the two that decide whether the split survives contact with a bad week.

Use it whenever a total has to be divided and more than one person has a view about how. The allocator is where the division is argued once, in writing, instead of repeatedly in meetings where whoever is unhappiest that week gets the increase.

Use it again on each row's own cadence. An allocation is not a decision you make in January; it is a decision you re-make on a schedule, with the previous version and its reasons in front of you.

Most budget splits exist as a number in a slide and a set of half-remembered reasons. That is fine until somebody wants more, at which point the conversation starts from nothing and the increase usually goes to whoever argues hardest rather than to wherever the money works best.

An allocator fixes that by making the argument the document. One row per destination, and in each row not only the share but what the share rests on: a baseline read from your own account over a named window, a contract that cannot be unwound, a test result, or a stated assumption somebody is willing to own.

The immediate benefit is that the next conversation starts from the last one. The longer-term benefit is that a year of allocators shows what the team has actually believed over time, and how often those beliefs were right, which is the only way a split gets better rather than merely different.

The basis column asks what kind of claim a share is. There are broadly four, and the damage comes from blending them silently.

Measured performance is a share argued from a number you can point at: this destination has produced results at this cost over this window. A strategic bet is a share argued from a belief about the future: a new channel, a new market, an audience you think is undervalued. A fixed commitment is money that is already promised — a contract, a sponsorship, a minimum spend. And a floor is a share somebody has insisted on for reasons that are political rather than empirical, which is a real category and worth naming honestly.

Keep them in separate rows. A bet hidden inside a performance line makes the line look worse than it is and makes the bet unaccountable; a commitment folded into a performance line makes a fixed cost look like a choice. Once they are apart, each can be judged on its own terms, and the total becomes readable: this much is bought on evidence, this much on belief, this much is not really available at all.

Shares are what people argue about. Floors and ceilings are what actually determine whether a split works.

A floor is the least a destination may hold and still be worth running. Below it, a line does not simply produce less — it produces noise: too few results to distinguish from chance, too little delivery to leave the platform's learning behind, and a cost per result that swings wildly enough to justify any conclusion anybody wants to reach. Writing the floor down forces the real choice, which is that a destination is either funded properly or not funded at all. A split spread thinly across nine destinations, each below its floor, is not diversification; it is nine experiments none of which will produce an answer.

A ceiling is the most a destination can absorb before the money stops buying anything new. Audiences saturate, lists are finite, inventory is capped, a market has only so many buyers at a price. Leave the ceiling blank where you genuinely do not know it — that is honest — but fill it where you do, because the most common way an allocator fails is by pouring the freed budget from one row into another row that was already full.

The reallocation rule is the column that makes the allocator usable after the first bad week.

A good rule has three parts: a condition, a size and a direction. If the cost per result stays above target for two consecutive weeks, move up to a fifth of this row's amount into the row with the best rolling result. That can be applied by whoever is looking, on a Monday, without a meeting.

The size limit matters as much as the condition. Unbounded reallocation produces a budget that chases last week everywhere, which reliably underperforms a split that moves slowly, because most week-to-week variation is noise rather than signal. A cap on how much may move at once is the cheapest protection against that, and it is much easier to agree in advance than in the middle of a bad fortnight.

Rules written under pressure get argued about. Rules written in advance get followed, which is the entire point.

Each row carries its own review cadence. A performance row might be weekly. A bet might be reviewed at the end of a quarter, because reviewing it sooner is just re-asking whether you still believe it. A commitment is reviewed when it can actually be changed.

Staggering the cadences is deliberate. An allocator that is entirely re-argued every month is an allocator nobody trusts, because the split never stays still long enough to produce evidence about itself. Fixing the cadence per row, in advance, is what lets a bet run long enough to be wrong in an informative way.

Six checks. Every row names a destination, a basis and an owner. Shares sum to one whole per currency. Every row has a floor and none is funded below it. Every row has a reallocation rule, written while nothing was going wrong. Every share names the evidence it rests on, or says plainly that it is an assumption. And the four bases sit in separate rows, so that the total can be read as what it is: a mixture of things you know, things you believe and things you have already promised.