Why a short cut from a long video fails

The workflow almost everybody lands on is the same. Finish the long video, scrub back through it for the good bit, top and tail it, crop it vertical, post it. It takes twenty minutes and it feels like free content.

Then it does nothing, and the conclusion people draw is that shorts do not work for their subject. Usually the subject is fine. The clip is the problem, and specifically the fact that it was never written to stand up on its own.

What is the long video giving the viewer that a clip cannot?

Context, and consent. Somebody watching your long video chose it. They read a title, looked at a thumbnail, decided the next quarter of an hour was worth it, and by the time they reach the good bit they have heard your voice for several minutes and know roughly what is going on.

A viewer meeting the clip may not have seen the long video. Check what the extract assumes: a term explained earlier, a question off camera or a demonstration already under way. Supply enough context for this version.

That is the whole failure, and it is not fixed by trimming harder.

Does the short need its own opening?

Yes, and it is the only part of this that is difficult.

The first sentence of a clip has to work for somebody who does not know the topic, the format or you. That almost never happens by accident somewhere in the middle of a long video, because the middle of a long video is written for people who have been there since the start.

Two ways to get one. Choose a moment that happens to begin with a complete statement, which is rare and worth looking for. Or film an opening for the clip while you are still set up, ten seconds straight down the lens that states the thing the clip is about. The second option is far more reliable, and it costs you a take.

Which pauses belong in the short?

Keep pauses that help the thought, joke or action land. A pause that only delays the start is a candidate to cut.

A long video is allowed to breathe, and it should. The pause before a difficult point is how a listener knows one is coming. The half second after a joke is where the joke happens. Take those out of a long video and you get something that sounds like a hostage reading a statement.

Tighten the run-up and remove repeated starts where they delay the point. Keep the breaths and pauses that carry thought, humour or emphasis. The shorter version still needs to sound like the person.

Push the tightening too far and the speaker can become accurate and characterless. Listen for whether the person survives the pace, not just whether every word fits.

Should it end, or should it loop?

Choose the ending the idea needs.

A loop can suit a repeating action or a line that returns naturally to the start. A demonstration, explanation or invitation may need a clear ending. Choose the structure for the piece.

A closing invitation takes time, so keep it relevant and concise. Judge it against the purpose of the video as well as retention; a useful enquiry may matter more than another loop.

How long should it be?

As long as the thing takes, and no house rule survives contact with a real audience.

A demonstration with three steps in it has a natural length, and cutting it to fit a number somebody read somewhere just removes a step. Equally, a fifteen second idea stretched to fill a minute is a fifteen second idea with padding on the end, and padding is visible.

The right length differs by who is watching. An audience of professionals looking for a specific technique will sit through more than a general audience being shown something for the first time. You get a read on which you have by posting the same idea at two lengths a few weeks apart and watching what happens. The weeks between the two posts change more than the length, so take the result as a lead worth testing, not proof.

A house style decides for you

This is worth understanding before you hire anybody.

A template can save work if it leaves room for the subject, voice and audience. Ask which decisions stay flexible and check whether the examples feel appropriate to your work.

In this studio the short is cut from the vertical sequence rather than from the horizontal edit, the opening is chosen from hooks that were filmed for the purpose and judged on how they held people, and the length is set per partner rather than by one internal rule. That is not a sophisticated system. It is the refusal to have a single answer.

Before your next batch of clips, look at the last five you posted and find the one that worked. Then ask whether you could write down what was different about its first two seconds. If you can, that is your next opening. If you cannot, that is the thing to go and find out.

Sources and evidence

What here is established about testing online and what is our own method. Platform behaviour was checked on 4 September 2026.

What is established. A before-and-after comparison run weeks apart is not a controlled experiment. Microsoft's account of large-scale online experimentation, and the peer-reviewed review of statistical challenges in online controlled experiments by Larsen and colleagues, both set out why: without randomisation, concurrent conditions and care around peeking at results early, a difference between two periods can come from many things besides the change being tested. YouTube's own guidance on understanding content performance is explicit that its performance data describes viewer response without by itself establishing why a change happened. That is the basis for treating a two-length comparison as a lead worth testing rather than proof.

Our method. Posting the same idea at two lengths a few weeks apart, and reading what comes back as a hypothesis rather than a verdict, is the practice here.

References.

• Microsoft Research. Online Experimentation at Microsoft. microsoft.com. Randomisation and proper design support causal claims. Platform experiments need power, guardrails and instrumentation.
• Larsen et al. Statistical Challenges in Online Controlled Experiments. tandfonline.com. Online tests have power, peeking, interference and metric pitfalls. Not specific to video platforms.
• YouTube. Understand your YouTube content performance. support.google.com. Performance data describes viewer response; it does not by itself establish why a change occurred. Official explanation remains observational.