How Long Should You Run a Meta Ad Before Judging It?

The most expensive habit in Meta advertising is impatience. An ad launches, the first day produces one expensive click, and by day three the budget has been changed, the audience narrowed and the creative swapped. Each of those edits restarts the process the system was in the middle of, which guarantees the instability that prompted the panic.

There is a defensible answer to how long an ad needs, and it is based on how the delivery system works rather than on a fixed number of days.

Why the first few days mean nothing

When an ad set is new, Meta’s delivery system has no information about who responds to it. It spends the early budget gathering that information, testing across different people, placements and times. This is the learning phase, and performance during it is deliberately unrepresentative.

Meta documents the threshold for exiting it as roughly fifty optimisation events in a week for an ad set. Until then, the cost per result fluctuates widely, and the numbers you are looking at describe an experiment in progress rather than an outcome.

Judging an ad during this period is like judging a recipe by tasting the ingredients separately.

The real measure is results, not days

The useful question is not how many days have passed but how many conversions have happened. Statistical confidence comes from volume, and two campaigns running for the same fortnight can produce entirely different levels of certainty.

As a working rule, you need somewhere between twenty and thirty conversions before the cost per result means much, and more than that before comparing two creatives against each other with any confidence. Below ten conversions, the variation between a good week and a bad one is larger than the difference you are trying to detect.

So the answer to “how long” is “long enough to produce that volume”, which for a small budget may be several weeks and for a larger one may be days.

A practical timeline

For most small business campaigns, a sensible schedule looks like this.

Days one to three. Check only that delivery has started, the ad is approved, and the conversion event is firing. Do not look at cost per result. Do not edit anything.

Days four to seven. Confirm spend is pacing and the tracking is recording conversions correctly. Still no structural changes. If something is genuinely broken, such as an ad disapproved or a pixel not firing, fix that.

Week two. Begin looking at cost per result as a direction rather than a verdict. Look at which creative is getting delivery, since the system concentrates spend on what works. Still avoid editing the ad set.

Weeks three and four. Now there is usually enough volume to make decisions. Pause clearly failing creatives, add new ones, and judge whether the cost per result is in the right range.

Underneath all of it, hold one principle: change one thing at a time, and give it time to settle before changing the next.

What does justify stopping early

Patience is not the same as ignoring obvious failure. Four things warrant intervention within the first week.

The ad is not being delivered at all, which usually means a rejection, a billing problem or an audience that is effectively empty. The conversion event is not firing, which means nothing is being measured. The traffic is clearly from the wrong place, which points at a location setting. Or the cost per click is wildly out of proportion to anything normal for your market, which suggests the creative is being ignored.

These are faults rather than results, and fixing them is not the same as reacting to early performance.

What editing actually costs you

Significant changes to an ad set push it back into learning. That includes changing the audience, changing the optimisation event, changing the creative substantially, and large budget changes.

The practical effect is that an account edited every few days never leaves the learning phase at all. Its costs remain permanently unstable, and the advertiser concludes the platform does not work, when the real issue is that it was never allowed to finish learning once.

If you need to make a substantial change, it is often cleaner to launch a new ad set alongside the existing one rather than to edit it mid-flight.

Creative fatigue, and the other end of the timeline

The opposite mistake is leaving a campaign untouched for months. Meta creative fatigues faster than most advertisers expect, particularly with a smaller audience where the same people see the ad repeatedly.

Watch frequency and cost per result together. Rising frequency with a rising cost per result is the signature of fatigue, and the answer is new creative rather than a new audience. Many accounts that are declared dead simply needed a refresh of what was being shown.

Planning for a regular supply of new creative, rather than a single batch at launch, is what keeps a working campaign working. That sits alongside the budget arithmetic in how much to spend on Meta Ads.

Deciding when a test has failed

Set the exit criteria before launching, in writing. Decide what a result is worth, how many results you need before judging, and what cost per result would make you stop. Then let the campaign run to that point without renegotiating with yourself at day four.

If the test reaches its volume and the cost is well above what the result is worth, stop. Before concluding the channel is wrong, though, check the two things that cause most false negatives: whether conversions are being tracked properly, and whether the landing page is losing the people the ads delivered. Both are covered in why Meta Ads get clicks but no conversions.

Running a test that gives you a real answer

A campaign judged too early tells you nothing and costs the same as one judged properly. The difference is planning: enough budget to reach a readable number of results, a defined period, a written threshold, and the discipline not to edit in between.

If you are unsure whether your campaign has run long enough to judge, or whether it is stuck in learning because of how it is structured, send us what you are running. We will tell you whether the result is real yet and what to change if it is not. Our Meta Ads management is built around testing properly rather than reacting weekly, and the wider framework is in our complete guide to Meta Ads.