Facebook ads usually need a learning phase of several days to a couple of weeks before their performance settles enough to judge, and the ads that truly work go on to run for months. The single most expensive mistake advertisers make is turning ads off during that early window, before the delivery system has learned who to show them to. Patience early, then a clear decision, then let winners run. This guide walks through the real timeline, backed by how long ads actually last in practice.
The short answer, stage by stage
An ad's life runs through predictable stages, and knowing which one you are in tells you whether to wait, judge, or scale.
- Days 0 to a few: the learning phase. The delivery system is testing who responds. Performance swings wildly and is not yet meaningful. Do not judge here, and do not keep editing, since every significant change restarts learning.
- After learning stabilizes: the judgment window. Once results steady and you have gathered enough of them, you can read the ad honestly. This is when you decide: scale, adjust, or cut.
- The long run: the winners. Ads that clear the bar can run for a very long time. The best ones run for months or years, quietly earning their budget every day.
The mistake is compressing this. People judge in the learning phase, panic at normal early swings, and kill ads that would have worked. The timeline is not a delay to endure, it is how the system finds your buyers.
What the learning phase actually is
When you launch an ad, the delivery system does not yet know who among billions of people will respond. It spends the early budget finding out, and while it learns, results are unstable and often worse than they will be once it settles. That instability is normal and temporary.
Meta's own guidance is that delivery is most reliable once an ad set gathers enough optimization events in a short window, which is why very small budgets and very narrow audiences struggle. If an ad cannot gather enough results quickly, it stays stuck in unstable learning and never settles. Two practical rules follow: give ads enough budget and audience to actually learn, and stop making significant edits to a live ad, because each one throws it back to the start.
How long winners actually run
Here is where real data reframes the whole question. People ask how long until an ad "works," but the more useful lens is how long working ads last, because that shows you what success looks like.
In the Hello Radio archive, a live archive of real ads from big brands refreshed every morning, the median active ad has been running a little over three weeks. Most ads simply do not last long, even from expert advertisers, because most ads are experiments that get cut. But the survivors run far longer. About 38 percent of active ads have run past a month, roughly 8 percent past ninety days, and a small elite past a year. The longest-running active ad right now has been live for more than 900 days.
The lesson is twofold. First, if most of your ads do not last, you are normal, that is how testing works. Second, when an ad clears the bar, the payoff is not a quick win, it is a durable one. The goal is not an ad that works this week, it is an ad you can leave running for months.
A realistic timeline to hold yourself to
Translate all of this into a schedule you can actually follow:
- Launch and leave it alone. Resist editing. Let the delivery system learn without resetting it.
- Wait for learning to stabilize and results to accumulate. Do not judge on a day or two of noise. Give it enough time and volume to gather a meaningful number of results.
- Make one clear decision. Once you can read it honestly, scale the winners, adjust the maybes, and cut the losers without agonizing.
- Let winners run and keep testing. A winning ad can run for months, so do not switch it off out of boredom. Meanwhile, keep testing new creative so the next winner is ready when this one fatigues.
Speed comes not from rushing the timeline but from testing smarter. The more you start from proven angles instead of blank guesses, the fewer expensive dead ends you sit through on the way to a winner.
What actually speeds things up
You cannot rush the learning phase itself, but you can shorten the road to a winner. A few things genuinely help, and none of them involve turning ads off faster.
Enough budget and audience to gather signal quickly. An ad that accumulates results faster clears learning faster and gives you a clean read sooner. Starving an ad of budget or squeezing it into a tiny audience does the opposite, keeping it stuck in unstable delivery for longer.
Fewer, stronger tests. Spreading a small budget across many ads means each one inches toward significance. Concentrating it on a few strong concepts gets each to a decision faster.
Starting from proven angles. The biggest time saver is not launching guesses. When you build from angles already shown to work in your market, more of your ads succeed and you spend less time sitting through dead ends.
Leaving winners alone. Every significant edit resets the clock. The fastest accounts protect their winners from tinkering and test new ideas as separate ads instead.
Notice what is not on this list: judging early. There is no version of "faster" that involves killing ads before they have learned. The speed comes from feeding the system better and testing smarter, not from losing patience.
Frequently asked questions
How long should I run a Facebook ad before deciding?
Long enough to clear the learning phase and gather a meaningful number of results, which usually means several days to a couple of weeks rather than one or two days. Judging during the early instability is the most common reason advertisers wrongly conclude an ad failed. Wait for results to steady, then make one clear decision.
Why are my Facebook ads doing worse at the start?
Because the delivery system is still learning who to show them to, so early performance swings and is often worse than it will be once it settles. This is normal and temporary. Do not panic or start editing, since significant changes restart the learning phase and reset the clock.
How long do successful Facebook ads run?
Far longer than most people expect. In the archive, winning ads routinely run for months, and the longest active ad has been live for more than 900 days. Only a small share of ads last past ninety days, so the ones that do are genuine winners worth leaving on. A durable ad, not a quick hit, is the real goal.
Does editing a Facebook ad restart the learning phase?
Significant edits generally do, which resets performance and costs you time and money. If an ad is doing well, leave it alone. If you want to test a change, run it as a new ad against the current one rather than editing the live winner, so you keep the proven ad running while you learn.
The takeaway
Facebook ads need a learning phase of days to a couple of weeks before you can judge them, and real winners run for months, not days. Give ads room to learn without editing them, judge only after results stabilize and accumulate, then let winners run while you test the next batch. Most ads will not last, and that is normal. The payoff for patience is a durable ad you can leave earning for a very long time.