Acme Watches' best ad is a creator unboxing. This month the team makes six iterations (three new hooks, a 9-second cut, a static pulled from a frame, and a second creator reading the same script) and two net-new concepts (a founder story and a side-by-side comparison). The iterations keep the winner fresh; the new concepts look for the next one.

Iterations are cheaper and win more often, because they start from something that already works, but they have a ceiling: the same idea reaching the same people. Net-new concepts fail more often, yet they are the only way to find a bigger winner or a new audience. Only iterating slowly fatigues an account; only making new concepts wastes budget.

Set a ratio and review it monthly. Many teams start with most output as iterations on winners and a steady share of new concepts, then shift the mix as results come in. Note which iterations worked, since that tells you which part of the winner mattered.