The brief said 15 seconds. It said 15 seconds because a blog post said 15 seconds, and that blog post said it because every other one did. The creator delivered exactly 15: a hook, a rushed demo, and a CTA that lands while she is still mid-sentence. It tested badly.

So the next brief said 30. The 30-second cut opened with eight seconds of b-roll before anyone said a word. That tested badly too.

Neither version failed on duration. Both failed because the number came first and the idea got stretched or squeezed to fit it.

Short answer: Video ad length is an output of retention. An ad can win at 10 seconds or at 45, as long as every second earns the next one. Brief the length your idea needs, then read the drop-off curve to see where viewers left, and cut to that point.

The takeaways

The 15-second convention is a measurement artifact. Meta's ThruPlay event counts a view at 15 continuous seconds or completion, whichever comes first, and that threshold became the number every length guide quotes.
ThruPlay is asymmetric. A 12-second ad has to be watched to the final frame to log one; a 60-second ad clears the same bar at 25% watched.
Placement caps are eligibility ceilings. Meta's specs allow 90 seconds in Reels and 120 in Stories, and almost nothing worth running goes near either.

Why does every video ad length guide say 15 seconds?

Because 15 seconds is where Meta starts counting. ThruPlay, the platform's standard video-view event, records a view when someone watches 15 continuous seconds or reaches the end, whichever comes first. The figure spread through every specs post because the platform turned it into a unit of measurement, and a measurement threshold is an easy thing to mistake for a creative rule.

Watch what that threshold does to two ads. A 12-second video has to be watched to the last frame before Meta logs a single ThruPlay. A 60-second video logs one at the quarter mark. Same event, very different bar. Optimise or bid on ThruPlay and the longer cut gets graded more generously, so a 12-second version and a 60-second version are never in a fair fight on that metric. Worth knowing before you decide the long one "held attention better".

What actually decides how long a video ad can run?

Retention decides it. A video ad can run as long as it keeps earning the next second. Attention in a paid feed is not handed over for a fixed duration and then withdrawn; it drains continuously, and every beat either refills it or does not. The practical test: if you cannot say what a given five-second block is doing for the viewer, that block is where people leave.

Which is why the same idea wins at very different lengths. A 10-second product reveal holds because there is nothing in it to get bored by. A 45-second founder story holds because it keeps opening loops. Both are fine. What dies is the 30-second version of a 12-second idea, padded with logo cards and drone shots, and the 15-second version of a 40-second idea, sped up until nobody can follow the argument.

How do you find the right length from your own data?

Read the drop-off curve on ads you have already run. Ads Manager reports video plays at 25%, 50%, 75% and 100% next to 3-second plays, and those five points sketch the shape of your retention. You are hunting for the cliff: the interval where the largest chunk of the audience walks.

A cliff between 3 seconds and 25% means the opening is the problem, and length is the wrong variable to touch (the hook is). A steady slope with a soft landing at 75% means the ad is roughly the right size and the tail is cheap to trim. A cliff between 50% and 75% is usually a dead middle: a demo repeating itself, a testimonial restating the hook. Cut the dead stretch, and the ad gets shorter as a consequence of the edit. That is the honest way to arrive at a number.

Do Meta's placement caps matter?

Only as ceilings. Meta's own placement specs allow 90 seconds in Reels and 120 seconds in Stories, and the feed permits far more than anyone should use. Those are eligibility limits: exceed them and the ad stops qualifying for the placement, which is a technical failure rather than a performance signal.

The placement question that does change the edit is pacing, not duration. A Reels viewer arrives mid-swipe, sound on, expecting motion. A feed viewer scrolls with sound off and reads the on-screen text. The same 25-second cut can hold on one surface and die on the other because the rhythm suits one and fights the other. Compose per placement, and the length question largely answers itself for each.

When is a longer video ad worth it?

When the viewer needs more before the offer makes sense. Awareness stage is the cleanest predictor here. Someone already comparing prices needs about eight seconds and a reason to act today. Someone who does not yet know the category exists needs the problem shown, the mechanism explained and the main objection answered, and that does not compress into 15 seconds without losing the argument.

Price does similar work. A €19 impulse buy sells in one beat. A €400 considered purchase usually cannot, and forcing it into the shortest format tends to buy a cheap click and no sale. So before you pick a duration, answer a smaller question: what does this viewer need to believe, and how many beats does that take? The seconds fall out of the beat count.

Where this stops being manual

Reading one retention curve and boarding one video is an evening's work. Doing it per concept, per awareness stage, per placement, every week, is the part that quietly stops happening.

That loop is what I built Adscalr for. It pulls the three public ad libraries daily and analyses competitor video frame by frame, longest-running first, for hook, pacing, emotional arc and CTA placement, so your pacing reference comes from ads that survived in market. On the creation side it produces motion storyboards composed for each native format, alongside static concepts and UGC briefs.

One limit I want to be exact about: Adscalr does not render or trim finished video, and a human approves everything before it ships. If turning research into structure is where your week goes, that is the ad-creation side of the product.

How long should a video ad be?