How kill criteria work
Kill criteria are written before an ad goes live and answer one question: what would have to happen for us to stop this? A usable criterion names three things — the metric that decides, the level at which the ad fails, and the minimum amount of evidence required before the rule can be applied at all. The last part is what stops a quiet first morning from killing a perfectly good ad.
Sensible criteria usually work in stages, because early events are more plentiful than late ones. An ad that almost nobody stops to watch can be judged on attention long before it has had a chance to produce sales. An ad that holds attention but produces expensive results is judged later, on cost per result. Setting the stage as well as the threshold is what makes the rule practical on a small budget.
Why kill criteria matter
They convert an emotional decision into an administrative one. Everyone who has run ads has kept a favourite creative alive on the argument that it needs more time, and has also killed a slow starter that would have worked. Rules written before the money was spent remove both mistakes, because the decision is made when nobody yet has a stake in the answer.
They also protect the budget in a specific way. Spend continues while a decision is postponed, so an unwritten rule has a running cost. On the modest budgets typical of local campaigns, a week of indecision is a meaningful part of the month.
Where kill criteria go wrong
The commonest fault is a threshold with no minimum evidence attached. Judging cost per result on a handful of events produces near-random verdicts, and ads get paused for having an unlucky first day.
The second is killing on the wrong metric. Cheap clicks that never become enquiries are not success, and expensive clicks that fill the diary are not failure. The rule should sit on the outcome the business actually sells, with earlier metrics used only as an early stage.
The third is applying the rule while the campaign is still relearning after a change. Delivery is unstable during the learning phase, so figures collected then are not evidence of anything, and pausing on them wastes the spend that would have taught you something.
How to act on it
Write the criteria into the launch note along with the target, so the test and its exit rule arrive together. Set the minimum evidence first — the spend or number of events below which you will not judge at all — then the threshold, then who is allowed to apply it.
Review the rules occasionally rather than treating them as permanent. Seasonality, a new offer or a changed landing page can move what an acceptable cost looks like, and criteria set for last quarter can quietly kill this quarter’s best creative. Pause rather than delete, and keep a short record of what was stopped and why; over a few cycles that log becomes part of your creative testing framework and stops the same idea being retried by accident.