Start from what a customer is worth instead. Take the gross profit on a typical customer, not the revenue, because you cannot spend revenue on advertising and the difference is frequently half. Multiply by how many times a customer buys in a realistic period, using your actual history rather than an assumption. That figure is the ceiling on what acquiring one is worth, and everything else follows from it.
Then work out how long you can wait to recover it, which for a business without reserves is the binding constraint rather than the total. A customer worth six hundred dollars over three years does not help you pay for advertising this month. The practical position for most first year businesses is that acquisition should roughly pay for itself on the first purchase, because spending against future value requires runway you probably do not have.
Treat the first budget as the cost of answering a question rather than as growth spending. The question is whether you can acquire a customer profitably through a given channel, and answering it requires enough volume to distinguish a result from noise. That usually means several hundred dollars a month sustained for two or three months rather than fifty dollars for a fortnight, which produces a number nobody can interpret.
Set the stopping rule before you start, because deciding afterward is how both common errors happen. Businesses quit at three weeks when the test needed eight, or continue for a year because stopping would mean admitting the spending was wasted. Write down what result would justify continuing, what result would mean stopping, and the date you will decide.
And remember what a percentage of revenue conceals. A business with strong referral flow may need almost no paid marketing, while one entering a competitive category may need considerably more than any benchmark suggests. The number that matters is your cost per acquired customer measured against what that customer is worth, and no industry average substitutes for it.
Separate the budget for testing from the budget for what already works, because mixing them produces neither. A proven channel deserves steady funding at a level you can sustain. A new one deserves a defined amount with an end date attached. Businesses that fund everything from one pool tend to starve the experiment and overfund the habit.