A small budget means fewer variables, not more
A large budget can drown noise in volume. A small one cannot. Test five variables at once and you end up with numbers nobody can explain. Fix four, move one — on a small budget it is the only approach that works.
Write the stop rule before you start
Decide in advance: how much, how long, what counts as a win, and what stops it immediately. A test with no stop rule turns into "let's give it a bit longer" until the budget is gone and there is still no conclusion.
A test without a stop rule spends money and returns a guess.
Small budgets test angles; large budgets test execution
Whether an angle works shows up for tens of dollars. Subtle differences in image and copy need a much larger sample. With limited money, the first is a far better return.
Leave one sentence behind after every test
Win or lose, write one line about what you learned. Three months later those lines are worth more than any single winning creative: they are the part you never have to test again.