· 20 min read
How Long Do Facebook Ads Run? Duration Is Three Numbers
Ad age, creative age and angle age are three different numbers, and only the first is visible for free. How to measure each one without quoting an average.
Search “how long do Facebook ads run” and you get a number. It is usually a round one, usually somewhere between a week and a month, and it is always built the same way: someone took a set of ads, measured the distance between each ad’s start date and the day it disappeared, and averaged the result. The number sounds authoritative. It also describes no actual ad.
Duration is worth measuring. It is the closest thing to a free profitability signal that exists in public ad data, because nobody keeps funding a loser indefinitely. But the useful question is not “what is the average” — it is “what is this advertiser’s kill discipline, and which of their ads has survived it.” Those are answerable, for free, in about twenty minutes a week. The average is not, and chasing it is how people end up copying a creative that a brand retired six months ago.
One Question, Three Different Numbers
The reason duration arguments go nowhere is that three different quantities hide behind the same phrase. They are measured differently, they mean different things, and only one of them is printed on the screen for you.
| Quantity | What it actually measures | Where it comes from | Free? |
|---|---|---|---|
| Ad age | Days since this specific ad record started delivering | The “Started running on” date on the card | Yes, directly |
| Creative age | How long this video or image has been in market, across every ad running it | Earliest start date among all cards sharing the asset | Only by reconstructing it |
| Angle age | How long the brand has been pushing this hook, claim or offer | Reading copy and offers across cards, over time | Only by logging it |
A card that says three days old can carry a video that has been in market since spring. A card that says two hundred days old can be a video nobody looked at since the week it launched. Both are ordinary, and both will mislead you if you read the date as “this creative is being tested” or “this creative is a proven winner.”
The distinction that matters most in practice is the middle one. Ad age is a property of an ad record, and ad records are cheap — a brand can create a new one by duplicating a campaign, changing an audience, or re-uploading the same file. Creative age is a property of an asset, and assets are expensive: they cost a shoot, an edit, a creator, or a week of someone’s attention. When you want to know whether a competitor believes in something, creative age is the number that answers it. Ad age is the number that is conveniently displayed.
Why No Average Is Worth Quoting
The population of live ads is not one distribution. It is a mixture of at least three regimes with completely different mechanics, and averaging across them produces an artifact.
Regime one: test-and-kill. Short-lived by design. A brand launches variations, reads them against a threshold, and turns off most of them within days. These ads are supposed to be brief; their short life is not a failure, it is the process working.
Regime two: workhorses. The ads that cleared the threshold and now carry budget. They live for weeks or months, and they get refreshed with new variants rather than replaced.
Regime three: infrastructure. Always-on retargeting, catalog feeds, and generic brand ads that run year-round regardless of performance, because they feed a pool or fill a slot.
Now put numbers on a hypothetical pair of advertisers with twenty-four live ads each. Both have the same mean ad age. Their actual behavior is nothing alike:
| Ad age band | Advertiser A | Advertiser B |
|---|---|---|
| Under 7 days | 4 | 18 |
| 7–14 days | 3 | 4 |
| 14–30 days | 4 | 2 |
| 30–60 days | 6 | 0 |
| 60–120 days | 5 | 0 |
| 120+ days | 2 | 0 |
Advertiser A has a portfolio: a steady churn of tests, a handful of funded workhorses, and two assets that have been in market for most of a year. Advertiser B is in a launch sprint and has not yet produced a survivor. A mean computed across both tells you nothing you can act on, and a mean computed within either one still hides the shape — which is the only part that predicts anything.
So when you see a published figure for average ad duration, three questions defuse it. What is the population — which advertisers, which categories, which countries? How was the end of an ad’s life observed — by the platform, or by a tool that started watching on some specific date and cannot see anything before it? And is the figure a mean, a median, or a percentile dressed up as a mean? In most cases the honest answer is “a tool’s own index, from its own start date, as a mean” — which is a description of the tool’s history, not of the market.
The Retention Rule That Sets Your Ceiling
Before any measurement method, you need to know what the library is willing to keep, because it decides whether duration is something you can look up or something you must record.
Meta documents this itself, and the asymmetry is large. For ordinary commercial ads that only deliver outside the EU and UK, the library shows the creative, the page, and the dates the ad ran — but there is no archive: once the ad stops, it leaves the default results. For ads delivered in the EU, the ad stays visible while active and is archived for one year after the last impression, with impression ranges and reach attached. UK-delivered ads get the same one-year archive plus reach breakdowns. Social issue, election and political ads are the outlier at seven years, which is why screenshots of spend data that circulate online are almost always political ads.
Three consequences follow, and each one changes how you should read a date.
Duration outside Europe is a forward-only measurement. If a competitor runs ads only in the US, there is no history to query. Whatever you did not record last month no longer exists. This is the single most common reason people conclude that a competitor “just launched” something that has actually been running for a year: they are looking at a re-upload, and the original is gone.
Europe creates a survivorship illusion. Because EU cards linger for up to a year after they stop, a card with a start date fourteen months ago is easy to find — and easy to misread as an ad that has been live for fourteen months. Check whether the card is active before you count it, and check which country’s results you are in. When in doubt, sort by active status first and treat anything inactive as a closed record with a known start date and an unknown end date.
Europe also gives you the only free volume data. The EU panel attaches impression ranges and reach to the ad, which turns a pure date into a rate: an ad that has been running sixty days with a wide impression range has been genuinely delivering, while a sixty-day ad in a narrow range may have been idling. Where that panel is unavailable, duration is all you have, and duration alone cannot separate a funded winner from a forgotten record.
The Four Windows and the Question Each One Answers
People filter by a duration window and then read the results as if the window were a quality grade. It is not. A window is a resolution: it selects a set of ads, and the set answers one specific question while remaining silent on the others.
| Window | What it selects | The question it answers | What it cannot tell you |
|---|---|---|---|
| 7 days | Ads launched this week | Is this new, and what is being tested right now? | Whether anything is working |
| 14 days | Ads that survived one full delivery cycle | Is the brand iterating on this, or leaving it alone? | Whether it is funded or merely paused-in-place |
| 30 days | Ads that crossed a monthly reporting and billing boundary | Is someone actively deciding to keep paying? | Whether the economics are actually good |
| 60 days | Ads that outlived at least one newer launch cycle from the same brand | Is this the control the account is built around? | Whether it was profitable the whole time |
The reasoning behind the boundaries is operational, not statistical. A week is the shortest interval at which a buyer typically reviews pacing and makes a keep-or-kill call, so surviving seven days means only that nobody intervened. Two weeks is the shortest interval that contains a full delivery cycle plus a stable second reading, which is why iteration — new variants on the same card, a swapped first frame — becomes visible there. Thirty days crosses a reporting boundary, which is when discretionary spend gets defended or cut. Sixty days crosses the brand’s own launch cadence: if they have shipped new creative since this ad started and this one is still running, it is being deliberately protected.
That last boundary is the most useful and the least used. It is also the one that produces the smallest sample. Filter a small advertiser to sixty days and you may get two cards, which is not a failure of the method — it is the method telling you that this advertiser has no established control yet, and that copying their “winner” means copying something a month old.
The interpretation ladder — what a thirty-day ad usually means, and the situations that break the rule — is the subject of the free competitor ads guide. This piece stays on measurement, because a correct interpretation applied to the wrong number is worse than no interpretation at all.
Why the Date Belongs to the Ad, Not the Video
Open any advertiser with a few dozen active ads and look for the same video appearing on two cards. It happens constantly: identical creative, identical copy, different start dates — sometimes weeks apart. That is the mechanic you have to design around. A start date belongs to one ad record. Duplicate the ad, rebuild the campaign, or re-upload the file after a pause, and you get a fresh record with a fresh date.
You can verify it in about two minutes on any brand that runs a lot of creative: find a distinctive visual, count how many cards carry it, and compare the dates. The spread tells you how much churn sits inside what looks like a single ad.
Two reading errors follow from ignoring this, and they pull in opposite directions.
Crediting a stale angle as fresh. A card three days old looks like a live experiment. If the same video has been in market for eight months across six cards, it is not an experiment — it is a retread of something the brand already knows, possibly already exhausted. Copying it means entering a fight you are eight months late to.
Dismissing a proven asset as a test. The reverse error: a video that has survived five months, relaunched into a new campaign last week, now reads as brand new. You skip it because “it hasn’t proven anything yet,” and you skip the single most validated asset on the page.
The fix is to key your notes on the asset, not the card. Record the visual, then the earliest date among every card carrying it, then the number of cards. The earliest date is a floor on creative age, never an exact figure, because there may be older records you cannot see. The card count is a second signal in its own right: one card means the asset is deployed once, six cards means the brand has rebuilt campaigns around it, which is what scale looks like from outside.
Same Asset, Several Countries: Two Patterns That Mean Opposite Things
Multi-market presence is usually read as strength, and sometimes it is. The start dates tell you which kind you are looking at.
Identical or near-identical start dates across markets. One launch, pushed everywhere at once. This is a campaign structure decision, often a duplicated campaign with translated text, and it says nothing about performance yet. Treat it as a single test that happens to be visible in six countries.
Staggered start dates, weeks or months apart. This is expansion, and expansion is a decision made after a result. An asset that goes live in one market, then a second six weeks later, then a third two months after that, is being rolled out against evidence. This pattern is much stronger than a simultaneous launch, and it is invisible unless you actually compare the dates country by country.
A market that appears after a gap elsewhere. When an asset dies in one market and shows up in another shortly after, that is reallocation, not growth. It usually means the first pool was exhausted or the first market’s economics failed, and the brand is testing whether a different audience tolerates the same creative.
There is also a mechanical effect worth naming, because it inflates duration in a way that has nothing to do with quality. Each market is a separate audience pool with its own delivery history, so the same asset faces independent repetition clocks — the ad fatigue guide covers how those clocks run. When one pool tires, the asset can keep running somewhere else. “Alive in six countries” frequently means “alive in one country and left running in five” — the aggregate looks robust while the actual performance is concentrated.
The practical read: measure duration per market, then count how many markets are under fourteen days, which is fresh expansion, versus over sixty, which is entrenchment. Those two counts describe the rollout strategy far better than a single global figure.
Build a Survival Table Instead of Chasing an Average
If you want a number that means something for a specific competitor, stop looking for their average and build their survival curve. It takes a fixed weekly log and about twenty minutes.
Six columns, one row per asset per check:
| Column | What to record |
|---|---|
| Check date | Same weekday every week |
| Advertiser and market | Page name plus the country filter you used |
| Asset key | A short description of the visual, not the ad ID |
| First seen | The earliest date you have ever observed for this asset |
| Cards | How many cards currently carry it |
| Status | Active this week, or gone |
After four or five weeks you can answer the question properly. Group assets by the week you first saw them, then count how many are still active at day 7, day 14, day 30 and day 60. That is a survival curve for that advertiser, and its shape is their kill discipline.
A hypothetical month makes the point. Twelve assets first observed in the same week, followed for sixty days:
| Day | Still active | Share |
|---|---|---|
| 0 | 12 | 100% |
| 7 | 11 | 92% |
| 14 | 8 | 67% |
| 30 | 4 | 33% |
| 60 | 2 | 17% |
Read the shape, not the endpoints. This advertiser barely kills anything in the first week, which means their short-lived ads are not evidence of anything — they may simply not have been reviewed yet. More than half of what they launch is gone by day fourteen, so surviving two weeks is a real filter here. And the two assets still alive at sixty days are the only two that cleared every cut, which makes them the highest-information ads on the page.
Now compare that to an advertiser whose curve drops to four by day seven. For them, a fourteen-day ad is already a strong signal, because their baseline is brutal. The same fourteen days means different things at the two accounts, and only the curve tells you which. This is why borrowed benchmarks fail: the number is identical and the meaning is not.
Two practical notes. Log assets you saw launch, not assets you found mid-life, because a curve built from ads that were already old when you first saw them is biased toward survivors. And keep the disappearance date, not just the status — knowing that something vanished between two checks is a decision timestamp, and it is the raw material for every “what did they stop doing” question later.
Three Ways a Long Run Means Nothing
Duration is evidence of continuation. It is not evidence of profit, and there are three common situations where a long-running ad proves nothing at all.
The retargeting net. A small audience, low delivery, months of runtime. These ads are structurally long-lived because they are shown to a warm pool rather than to a cold market. If a card has been running for a hundred days and points at a homepage or a collection rather than a specific product, treat it as infrastructure rather than a winner.
The evergreen brand ad. Some accounts keep a generic “shop the range” creative running year-round because it feeds retargeting and fills cheap placements. There is no offer, no product focus, and no relationship between runtime and performance. Long duration here is a scheduling choice.
The orphan. Plenty of ads keep running after the page behind them broke: the product is out of stock, the price is stale, the landing page 404s. Nobody paused it because nobody looked. Follow the link before you credit the runtime, every single time — it costs fifteen seconds and invalidates a surprising share of long-running ads.
There is a fourth case in the opposite direction, and it matters just as much: ads that are supposed to die. Sale and seasonal creative is built to run for the length of the promotion. A nine-day ad that carried a brand’s entire peak week looks identical, on the date field, to a nine-day test that failed. Judging it by runtime is a category error. During promo windows, read launch timing and offer terms instead of duration, and treat the evergreen ad that kept running alongside the sale as the real control.
What Duration Cannot Tell You
Keep the boundaries explicit, because the temptation to convert a date into a verdict is strong.
Whether delivery was continuous. A start date marks the beginning of a record, not an unbroken run. An ad can be paused and resumed, and the original start date stays. Where the EU panel is available, impression ranges give you a rough check on whether the ad has been actually delivering; where it is not, a ninety-day-old ad may have spent sixty of those days switched off.
Spend, revenue or profit. Outside the EU and UK there is no delivery data at all — no ranges, no reach, no impressions. Inside, you get binned ranges rather than precise figures. No platform publishes advertiser-side performance, so any tool quoting a competitor’s profit is inferring it.
What was attempted and rejected. You see what survived review. Ads removed for policy reasons never become part of the record, so the visible population is filtered before you start reading it.
Organic and other channels. The library covers paid social on Meta’s surfaces. A competitor’s best-performing channel may not appear in it at all, which makes any conclusion about “their strategy” partial by construction.
Verified 2026 · 20% off Starter, Pro and Business for your first 3 months
The manual method has one honest limit, and it is arithmetic rather than skill: a survival curve needs repeated observation, and repeated observation across many advertisers does not fit in a spreadsheet. TrendTrack exists for that layer — 10M+ Meta ads and 300M+ total ads across 1.5M+ Shopify stores, with each advertiser’s ads tracked over time, so a start date stays attached to a history instead of a single snapshot. The free Chrome extension shows a slice of it without an account, and the extension guide covers what each panel reads from. When you do subscribe, TrendTrack promo code is WINNING - 20% off all plans for the first 3 months, after which standard pricing applies.
FAQ
How long do Facebook ads usually run?
There is no single figure worth quoting, because live ads are a mixture of short tests, funded workhorses and always-on infrastructure, and averaging across those regimes produces a number that describes no real ad. The answerable version of the question is advertiser-specific, and you get it by logging a set of their assets weekly and counting how many survive to day 7, 14, 30 and 60.
Where do I find how long an ad has been running?
Each active ad card in the Meta Ad Library prints a “Started running on” date next to its library ID. Count the calendar days from that date to today and you have the ad’s age. Treat it as the age of that one ad record rather than of the creative, since a duplicated or re-uploaded ad starts a fresh record with a new date.
Does a long-running ad mean it is profitable?
No. Duration is evidence of continuation, not of profit, and three common cases break the inference: retargeting ads that run for months against a small warm pool, evergreen brand ads kept live to feed those pools, and abandoned ads whose landing page is dead or out of stock. Long runtime is a reason to look closer, never a conclusion on its own.
Why can I still see ads that stopped running months ago?
Because retention depends on where the ad was delivered. Meta archives ads delivered in the EU and UK for one year after their last impression, and social issue, election and political ads for seven years. Ordinary commercial ads shown only outside those regions leave the default results once they stop, which is why the archive feels inconsistent from one advertiser to the next.
The start date is three days old, but I have seen this video for months. Why?
Because the date belongs to the ad record, not to the asset. Brands duplicate campaigns, rebuild ad sets and re-upload the same file, and each version gets its own start date. To estimate creative age, find every card carrying that visual and take the earliest date among them as a floor, then count how many cards exist as a separate signal of how hard the asset is being pushed.
Which duration window should I filter for?
Choose the window from the decision you are making. Seven days finds what is being launched right now, fourteen shows what survived a full delivery cycle, thirty identifies ads someone is actively funding, and sixty finds the control the account is built around. A longer window is not more accurate, only narrower, and on a small advertiser it can leave you with two cards to read.
Final Word
The question “how long do Facebook ads run” has a satisfying answer only if you narrow it to one advertiser and one asset. Once you do, the number stops being trivia and becomes a reading of how that team operates: how fast they kill, how long they defend, whether a new start date means a new idea or an old file with a fresh coat of paint.
So skip the average. Pick five competitors, log their live ads by asset every week for a month, and count what survives. You will end up with a curve nobody can sell you, calibrated to your own category, and you will read a start date correctly for the rest of your research life. If you want the interpretation layer on top of that — which survivors are worth studying and what to take from them — the ad library search guide covers query construction, the hook research guide covers reading openers at the angle level, and the competitor budget guide explains what the aggregated numbers can and cannot prove.