How much should you spend before judging a Meta ad?

Spend enough to clear two gates before you judge a Meta ad: a real statistical sample AND the learning phase, scaled to your price point. A common working floor is roughly 1,000 impressions and about 50 to 100 dollars, but the honest bar is the spend it takes to see the number of conversions you would need to trust the result at your average order value. Judge before that and you are reacting to noise, not signal.

A single dollar figure ("spend 50 then decide") is the wrong mental model, because it ignores what you sell and how Meta delivers early.

The two gates

An ad is not judgeable until both are true.

Gate one: a real sample. A handful of impressions and one or two purchases can swing your cost per result wildly. You want enough clicks and conversions that the result would roughly hold if you ran the same ad again. Roughly 1,000 impressions and 50 to 100 dollars is a common starting floor, but it is a floor, not a finish line.

Gate two: the learning phase. During learning, Meta is still working out who to show the ad to, and cost per result is noisy and usually worse than it will settle to. A verdict made mid-learning is a verdict on unstable delivery. Let the ad exit learning, or get close, before you read the numbers as signal. More on that in the Meta learning phase, explained.

Scale the bar to your average order value

This is the part the flat "spend 50 dollars" rule misses. A 40-dollar product and a 400-dollar product do not deserve the same judging bar. Higher-priced products convert less often per click, so you need more spend before one purchase (or zero purchases) actually means something.

| Average order value | Rough spend before you can judge | Why |

|---|---|---|

| Low (under ~50) | ~50 to 100 | conversions come quickly, sample fills fast |

| Mid (~50 to 150) | ~150 to 300 | fewer conversions per click, needs more volume |

| High (~150+) | ~300+ | rare conversions, one sale is noise until spend is real |

These are directional, not laws. The real test is always: do I have enough conversions that this number would repeat?

Materiality: did the ad even get a fair chance?

One more check before you judge: did the ad get real delivery? If it spent only a tiny slice of its ad set budget, it never had a fair shot, so its bad number is the ad set starving it, not the ad failing. Do not judge an ad that never got enough delivery to be judged.

What to do while you wait

Do not sit and stare at an ad that has not cleared the gates. Make sure the ad set is not starving it, keep the creative pipeline moving so you always have the next test ready (creative testing framework), and resist editing the live ad, since editing resets its learning. Once both gates are cleared and the ad is clearly below target on a rolling window, then you have earned a decision: when to kill a Meta ad vs give it more time.

The whole point is patience with a rule attached: enough spend to form signal, scaled to your price, out of learning, with fair delivery. That is when a Meta ad is finally judgeable. For where this sits in the full decision loop, see how to decide what to change in your Meta ads.