Guide
The hardest call in paid media isn't launching — it's pulling the plug. Kill too early and you bury an ad that hadn't gathered signal yet. Kill too late and you've funded a loser out of hope. The line between the two is data, not feeling.
The reason good marketers keep losers alive is psychological, not analytical. You spent money making the ad, you believe in the concept, and killing it feels like admitting the spend was wasted. So you give it "one more day," then another. But the budget already spent is gone whether you keep running or not — the only question that matters is whether the next dollar does better in this ad or a different one. Framed that way, the answer is usually obvious well before you act on it.
The opposite mistake is pulling an ad on a bad afternoon. Early performance is noisy — a few hundred impressions tell you almost nothing, and the algorithm is still finding the audience. Give a creative enough volume to be a real sample: a common rule of thumb is a few thousand impressions, or spend roughly equal to your target cost per result, before you trust the number. Below that threshold you're reacting to variance. The goal is to kill on evidence, not on the first dip.
Once you have signal, look at three things. First, the hook rate — if almost no one watches past the first few seconds, the ad is dead at the door and no downstream tweak saves it. Second, cost per result well above your break-even after meaningful spend, with no improvement as the algorithm settles. Third, the trend: a creative that started strong and is steadily decaying is showing ad fatigue and won't recover on its own. A flat-bad line and a falling-from-good line both mean stop — for different reasons.
Killing an ad isn't a loss if you learn from it. Note what the losing cut had in common — the hook type, the format, the pacing — so your next batch doesn't repeat it. This is where a naming and versioning system pays off: when you can see that every ad with a slow, branded opener died, you stop making them. A kill is only wasted if it teaches you nothing.
The cheapest ad to kill is the one you never launched. You can't predict conversions in advance, but you can predict whether creative holds attention — and the ads that fail on attention are the ones that die fastest and clearest in-platform. Pre-testing scores your creative for predicted attention before launch, so the cuts that reach live spend already cleared a screen. You still have to make kill calls — pre-testing predicts attention, not conversions, and it never replaces live data — but you make them between real contenders instead of funding creative that never had a hook. Fewer obvious losers reach your budget, and the kill decisions that remain are the interesting ones.
Upload your creative and see its predicted attention before you spend a dollar deciding — first analysis free.
Pre-Test Your Creative