What is Kill Criteria?

Kill criteria are pre-defined rules for when to turn off an ad or ad set — for example a spend threshold with no purchases, or a CPA ceiling — decided before the test launches.

How it works

Deciding the exit rule in advance removes emotion from creative testing: without it, advertisers either kill ads too early on thin data or keep losers alive out of sunk-cost hope. Good kill criteria are tied to your economics — typically a multiple of your target CPA or AOV in spend without the required result. They turn testing into a repeatable system where every creative gets the same fair, finite trial.

What to watch

• Good: every test launches with a written spend limit and result threshold before it goes live
• Bad: ads get paused after a few hours based on gut feel before they have meaningful data
• Bad: losing ads keep spending for weeks because 'they might turn around'

In practice

Strict kill criteria only work if you have enough creatives queued that killing a loser costs you nothing emotionally or operationally. Generating test batches inside Claude with Polaris means there is always a next challenger ready, which makes it easy to actually enforce the rule. See the full ecommerce ads glossary, or put it into practice with the Polaris AI UGC ad generator.