A founder looking for a marketing budget usually finds a percentage. Spend a share of revenue, a share of the round or whatever a benchmark says companies of a certain size spend. The number feels safe because someone else picked it. For an early startup it is almost always the wrong place to start, because the percentage answers a question the company is not asking.
Percentages describe companies that already know what works.
Revenue-based rules come from established businesses. Those companies have channels that produce customers at a known cost, and the budget question for them is how much more to put into a machine that already runs. A share of revenue is a reasonable way to size that.
An early startup has no machine. It may have little revenue, which makes a share of it close to nothing, or a fresh round, which makes a share of it far more than marketing can usefully spend before anyone knows what works. In both cases the percentage sets the size of the budget without saying what the money is for.
A good budget buys an answer.
The better question is not how much to spend. It is what the company needs to learn next, and what the smallest amount is that will teach it.
Early marketing is a series of questions. Will this buyer respond to this message? Does this channel produce qualified conversations at a cost the business can carry? Will people who arrive from search fill in the form? Each question has a price: the spend and the time it takes to get a clear yes or no. A good budget is the sum of those prices for the questions that matter this quarter, and not much more.
That turns the budget from a number into a plan. Each line has a question, a spend, a window and the result that would count as an answer. When the window closes, the line either earns more money or stops.
Ninety days is long enough to learn and short enough to stop.
Most early tests should fit inside a quarter. Ninety days is long enough for a paid channel to produce readable results, for a handful of outbound sequences to run their course, or for a new home page to collect enough visits to say something. It is short enough that a failing test does not quietly become a permanent line item.
Some work takes longer. Search and content usually need several quarters before they show much. That is fine, as long as the budget says so in advance and the first quarter's question is a smaller one: are the right pages getting indexed, are the first visitors the right people. A long bet still needs a short checkpoint.
The budget includes things that are not ad spend.
Founders often count only the money that goes to platforms. A real budget also counts the tools, the contractor or agency time, the design and writing, and the founder's own hours, which are the most expensive input most startups have and the one least often written down.
Counting all of it changes decisions. A channel that looks cheap in ad spend can be expensive once the hours to run it are included. An agency that looks expensive can be cheaper than a founder spending two days a week on work they do slowly. For one common stage, see What a seed-stage SaaS startup budgets for marketing.
Set it from the questions, then check it against runway.
Start with the list of questions for the next ninety days, in order of how much each answer is worth. Price each one. Add them up. Then compare the total with the runway and with what the company will need to spend once a channel works, because a successful test is only useful if there is money to scale it.
If the total is too high, cut questions from the bottom of the list, not money from every line. A test funded at half the amount it needs usually produces an unclear answer, which costs the full amount in time and teaches nothing. Fewer, properly funded questions beat many underfunded ones.
The percentage still has a use later. Once a channel works and the company knows what a customer costs, a share of revenue is a sensible way to grow the budget with the business. The wider budget question sits in How much a startup should spend on marketing.
List the three questions marketing has to answer this quarter, and fund those first.
Tell us what you're building.