How long should a promo video be?

The question is usually asked once, at the end, when the video is already three times too long and someone has to decide what to cut. Asked at the start it is a much better question, because the answer sets the number of scenes, and the number of scenes sets what can go in them.

What the completion data says

Two large samples, measured differently, point the same way.

Vidyard’s benchmark covers nearly a million videos hosted on its platform, 943,305 of them created during 2024. Of videos under a minute, 65% of viewers stay engaged to the end. Over twenty minutes, only 20% do.

Wistia’s 2026 State of Video report draws on over 13 million videos and 79 million hours of viewing. Its figure for videos under a minute is a 52% average engagement rate, with product videos at 50% and customer testimonials at 46%.

Those two numbers are not the same measurement and should not be added together or averaged. One is the share of viewers who reach the end, the other is the share of the video an average viewer watches. What they agree on is the shape: the curve falls off fast and it falls off early.

Wistia’s own caveat is the honest counterweight, and worth repeating. A ten minute video with a lower engagement rate still produces more total watch time than a one minute video with a higher one. If the thing you need is minutes of attention from a small number of people, short is the wrong optimisation.

Length by placement

The table at the foot of this post sets six placements against the target length and scene budget the planner uses, and against what each platform publishes for itself.

A few things stand out when you put them side by side.

The platforms are permissive and the audiences are not. TikTok allows ten minutes in feed and no limit at all on Spark Ads. LinkedIn allows thirty minutes. YouTube’s skippable in-stream format has no maximum. None of that is a recommendation, it is a ceiling, and the only ad format with a genuinely short published limit is the YouTube bumper at five to six seconds.

The only hard clock is the skip button. YouTube’s skippable in-stream ads become skippable after five seconds. That is the real length constraint on a pre-roll: you have five guaranteed seconds and everything after them is optional viewing you have to earn.

TikTok publishes timing guidance rather than a duration. Its creative best-practice page asks for the content proposition in the first three seconds and the hook prioritised inside the first six, and suggests keeping on-screen text to five to ten words a second. That is more useful than a length. The frame shape matters as much: the aspect ratio calculator will tell you what a given source crops to before you commit to a placement.

One gap worth stating plainly: Meta hosts an article on Reels ad specifications, but its help centre renders entirely in the browser and returned no readable text on any of the ten attempts made on 1 September 2026. Rather than repeat a figure from a marketing blog, this article quotes nothing for Reels. Treat 15 to 30 seconds as a working assumption and check Meta’s own page before you buy. Where text can safely sit inside that frame is a separate question and worse documented still, which the safe zone article goes through platform by platform.

Purpose changes the length

Placement sets the ceiling. Purpose sets where inside it you land.

An awareness video is short by nature. It has one job, which is to make a stranger recognise the name later. Ten to fifteen seconds, one claim, no explanation.

A demonstration cannot be ten seconds and should not try. It needs the problem, the action and the result, which is three beats minimum, and each beat needs long enough to be read. Thirty seconds is the floor and sixty is comfortable.

A launch video sits in between, usually because it is doing two jobs at once, and it is the one most likely to sprawl. If the launch video is over a minute, look for the demonstration hiding inside it and cut that out as its own video.

The scene budget

Length is not a slider. It is a scene count, and the planner works to three fixed budgets.

Length	Target seconds	Scenes	Seconds a scene	Narration a scene
Short	10	3 to 6	2.5 to 4	5 to 8 words
Standard	18	4 to 9	2.5 to 4	5 to 8 words
Long	30	6 to 12	2.5 to 4	5 to 8 words

Read that table the other way round and it becomes a planning tool. If you have six things to say, you are making a standard video, and a short one will drop two of them. If you have twelve, you are making a long one and it will still be tight.

The 2.5 to 4 second scene is where most of the discipline lives. It is long enough to read one line and see one image, and not long enough for two of either. A scene that needs six seconds is usually two scenes.

Creator-style ads are stricter again. A UGC video is capped under thirty seconds by design, with at most three talking scenes and non-talking scenes held to 2.5 to 3 seconds. The UGC script guide covers what goes in those beats.

Narration words per second

Five to eight words a scene sounds miserly until you time it. At a normal speaking pace of around 150 words a minute, eight words is a little over three seconds, which is the whole scene.

This is the single most common reason a planned fifteen second video comes out at twenty six. Narration stretches the scene it sits on, so an extra clause in every line adds up across six scenes. Write the lines first, read them aloud against a clock, then decide the length.

When longer is right

Three cases. A video living on a page a visitor chose to open, where they have already committed. A demonstration where the thing being shown is genuinely the point. And a video going into an email or a proposal, where the audience is small, specific and already interested.

In all three the audience arrived on purpose. Feed video has no such luxury.

Plan the fifteen second version first

Whatever length you end up making, write the fifteen second cut first. Three or four scenes, one claim, one proof, one close.

It is a better brief than a long one, because it forces the single most important sentence to the front, and everything longer can be built by adding to it. Going the other way, cutting sixty seconds down to fifteen, almost always leaves the best line stranded at the end where nobody reaches it. If the fifteen second version does not work, the sixty second one is not going to be saved by its extra forty five.