When to Kill an Ad Set

Knowing when to kill an ad set is mostly about not doing it too early: giving it enough data to judge honestly, then cutting it once it has genuinely proven it cannot make you money.

The most expensive habit in a young account is killing things on a bad morning. An ad set in its learning phase is supposed to look jumpy, one day cheap, the next day awful, because it is still working out who to show your ads to, and a decision made on day two is a decision made on noise. So the first rule of killing an ad set is patience: let it get through learning and gather enough of your optimization event that the number in front of you is a real signal and not a coin toss. Judging an ad set before it has spent a meaningful multiple of your target cost per result is judging a story you have only read the first page of.

Once it has had that fair shot, the decision gets simpler and colder. If an ad set has left learning, spent enough to be believed, and is still coming in above what a result can cost you and stay profitable, that is your answer, and no amount of waiting is likely to rescue it. Consistently unprofitable after a fair test is the signal to cut. The mistake in the other direction, letting a proven loser run because you are attached to it or hoping it turns, is just a slower version of the same waste.

Getting out of learning is not the same as being profitable, and that cuts both ways. An ad set can finish learning and still lose money, and an ad set stuck in learning limited can still be quietly making you money, so kill on the economics, on whether it earns, not on the status label Meta shows. That judgement about your own numbers is yours to make, not Meta's.

Do not kill on a bad day and do not nurse a proven loser, judge it on whether it earns after a fair test, and nothing else.