When to Kill vs Scale a Meta Ad: A Decision Framework

Most bad account decisions are made on a bad day or a lucky one: a buyer kills a good ad after two soft days, or floods budget into a fluke. The cure is to judge every ad against the same three questions in the same order, every time. Volume, then trend, then standing.

Is there enough volume to decide at all?

A 0.8 ROAS on 12 dollars tells you nothing; a cost per result that doubled on four clicks tells you nothing. You need enough events behind a number for it to be signal, not luck. Meta's own systems work on a similar threshold: an ad set is expected to stabilise only after about 50 results in the week following its last significant edit, and below that it is flagged "learning limited" and delivery stays jumpy (Meta Business Help). If your ad has not cleared a comparable bar, the only correct action is to wait. This discipline prevents most mistakes; it has its own article: decide only on spend too small to judge.

What is the trend, not the snapshot?

One day is weather; a week is climate. Read the direction over a rolling window. An ad at 1.6 ROAS falling for five straight days is a different decision from one at 1.6 climbing out of a dip. Kill candidates show a sustained decline: falling click-through, rising cost per result, rising frequency, across enough days that it is clearly not noise. If the decline traces to the audience wearing out rather than the ad being weak, the answer may be a refresh, not a kill: Meta ad creative fatigue.

How does it stand against your own ads?

An ad is not good or bad in absolute terms, only relative to what else could spend that money. Rank your active ads on the same objective. The consistent bottom of the list, on real volume, is where budget is wasted, even if its raw number looks acceptable. "It is still profitable" is not a reason to keep an ad that sits last of ten; that budget earns more in your top three.

The decision, and one delivery caveat

• Kill when an ad is below target, on enough volume, flat or worsening, and bottom of the pack. Killing frees budget; it is not a failure.
• Scale when an ad is above target, on real volume, stable or improving, and near the top. Raise budget in steps, because a large budget jump is a significant edit that can reset the learning phase (Meta Business Help); Meta notes small changes are usually fine but large ones may re-trigger learning.
• Leave alone the middle. Most ads live here, and constant tinkering just resets learning and costs you more than it earns.

Run the three checks the same way every time and your account stops swinging on noise. The level you read these metrics at matters too, because the same ad looks different at ad, ad-set, and campaign level: which metrics to read at each level.