A business that makes money has to decide what happens to it. That decision, repeated, does more to determine what the business becomes than almost anything in the operating plan — and in most small businesses it is a residue rather than a decision. Money accumulates, something urgent appears, the money goes there.

The formal name for doing it deliberately is capital allocation, and there are only a handful of places the money can go.

The options

Back into the existing business. More of what already works: more marketing into a channel that converts, another person on the team that is the constraint. The return here is the most estimable of any option, because the business already has evidence.

Into something new. A new service, a new market, a product. Higher variance, and the honest expected return is usually lower than it feels, because new things have no track record and the estimate is coming from enthusiasm.

Into reserves. It looks like doing nothing and it is not. Cash buys the ability to survive a bad quarter and to act when something cheap appears, both of which have real value that never shows up as a return.

Out to the owner. Also a legitimate allocation. The business exists to produce a life, and the owner who reinvests everything for a decade has made a choice, whether or not they ever framed it as one.

Into paying down debt. A guaranteed return equal to the interest rate, which is more than most of the alternatives can honestly claim.

The only question that matters

Every one of those is defensible. What makes the decision tractable is that they are mutually exclusive: a dollar spent on one is a dollar not available for the others, which is opportunity cost in its purest form.

So the question is never "is this a good use of the money". Almost everything is a good use of the money considered alone. The question is "is this the best available use of the money, given the others" — and asking it that way kills a great many proposals that survive the first version.

The most common failure is not a bad allocation. It is allocating by urgency, where the money goes to whatever is loudest this month. Over a few years that produces a business shaped by its interruptions.

What the returns actually look like

Reinvesting in the existing business runs into diminishing returns faster than anyone expects. The first $5,000 into a working channel might return several times over; the tenth $5,000 into the same channel usually does not, because the cheap attention was bought first. This is why businesses that reinvest mechanically into what worked last year see the return decay, and why the decay is often read as a marketing failure rather than as the shape of the curve.

Reserves are systematically undervalued because their return is invisible. The value of having cash is realized on the day something goes wrong or something good becomes available cheaply, and on every other day it looks like idle money. That is optionality, and it is worth paying for precisely because you cannot know in advance which day you will need it.

And the returns compound, which is the argument for caring about the decision at all. A business that allocates moderately well for ten years and one that allocates by urgency do not end up 10% apart, because each year's allocation determines the size of the next year's decision. Compounding applies to judgment as much as to money.

Doing it deliberately without a finance function

Three things, none of which require a CFO.

Decide on a schedule, not on a trigger. Quarterly, with the surplus counted and the options written down. Deciding when an opportunity appears means deciding when the comparison set is empty.

Split the decision before you make it. A rough policy — a share to reserves, a share to the owner, a share to reinvestment — settles most of the allocation in advance and leaves argument only about the remainder. It is worse than a perfect case-by-case judgment and much better than what happens without it.

Write down what you expected. Not for accountability; for calibration. The purpose of recording that a $10,000 marketing spend was expected to return $30,000 is that in six months you find out whether your estimates are any good, and that is the input to every allocation after it.