How long should a marketing video be?

The curve has a bump in it

The dip at one to three minutes and the recovery at three to five is the most useful thing in Wistia's benchmark table, and it is the part nobody quotes. A one to three minute video is usually a product overview: long enough to need an argument, short enough that nobody built one. A three to five minute video is usually a tutorial or a walkthrough, where the viewer arrived wanting a specific thing and the video is the fastest way to get it.

In other words the bracket is not measuring duration. It is measuring how well the intent of the viewer and the intent of the video match, and duration happens to correlate with that. Which is why cutting a bad three minute video down to ninety seconds usually produces a bad ninety second video.

Wistia's own 2026 report, drawn from over 13 million videos and 79 million hours of viewing, states the headline more bluntly: the shorter the video, the higher the engagement rate, with videos under a minute averaging 52 per cent. Both things are true. The trend is downward and the table is lumpy, and a marketer who only knows the trend will make the wrong cut.

Cutting a bad three minute video down to ninety seconds usually produces a bad ninety second video.

Engagement rate and completion are different questions

Wistia reports engagement rate: the share of the video an average viewer watches. Vidyard reports something closer to completion: the share of viewers still there at the end. Across 943,305 videos created in 2024, Vidyard found 65 per cent of viewers stay to the end of a video under a minute, and 20 per cent for videos over twenty minutes.

Those two metrics move together but they answer different briefs. Engagement rate tells you whether the video holds attention while it plays. Completion tells you whether the payload at the end ever lands. If your call to action is the last eight seconds, completion is the only number that matters and a 38 per cent engagement rate on a ten minute video means almost nobody heard it.

The practical consequence is unglamorous: put the ask where the audience still is. A video whose closing frame carries the offer is betting everything on the tail of a distribution that both of these datasets describe as thin.

People are making more long video, not less

Here is the finding that embarrasses the short-form consensus. In Vidyard's 2024 data, videos over twenty minutes grew by 420 per cent year on year. Videos under three minutes grew by about 30 per cent, and the three to twenty minute band by around 60. Total video creation rose 88 per cent, and the average person went from making roughly one video a month to nearly one every ten days.

So the fastest growing category is the one with the worst completion rate. That is not irrational. Those are demos, recorded meetings, webinars and training, and their job is not reach. A twenty minute recording watched to 20 per cent by eleven people at the right account is a better afternoon than a fifteen second clip watched to the end by nine thousand strangers.

Run the arithmetic on Wistia's own table and it makes the same point from the other side: 25 per cent of a forty five minute video is over eleven times the watch time of 48 per cent of a two minute one. Rate is not volume. Choosing a length by comparing rates alone is how a team ends up making nothing but clips and wondering why the pipeline is dry.

What actually decides the number

Three questions, in this order. Where will it play, and is the sound on? What has to be true in the viewer's head by the end? How much of that is already true before they press play?

A cold audience needs the problem established, so a 15 second cut can only carry one claim. A warm audience on a pricing page already accepts the problem, so 30 seconds can spend all of itself on proof. A customer in an onboarding email accepts both and only wants the steps, which is why the three to five minute bracket holds up.

Notice that none of those questions is about the video. Length is an output. When someone asks for a sixty second film, the honest first reply is a question about where it plays, and if the answer is a muted feed then sixty seconds was never the brief.

What to do about it

1. Write the placement and the sound state at the top of the brief, before anyone proposes a duration.
2. Move the call to action to wherever the audience still is, which for anything over five minutes is not the last frame.
3. Stop comparing a clip and a webinar on engagement rate. Compare them on total watch time and on what each one is for.