Why YouTube videos feel three times longer than they need to be

You clicked on a video that promised to explain something. Four minutes later it has told you that this is a topic many people struggle with, that it took the presenter years to understand, and that the answer is going to surprise you. It has not yet started.

That feeling is not impatience and it is not a shortened attention span. It is an accurate reading of the ratio, and it can be measured.

Across 3,486 videos we have measured, the median padding share is 40%. Not the worst offenders — the median. In a typical explanatory video on YouTube, around two minutes in five is carrying the video forward without adding to it.

The padding is not where you are looking

Ask people where the waste is and almost everyone says the same two things: the intro and the sponsor read.

The sponsor segment is the easier one to dismiss. It is typically sixty to ninety seconds in a video running ten minutes or more — call it 10% at the high end, and it is clearly signposted, easy to skip, and the reason the video exists at all. Removing every sponsor read from every video we have measured would move the median padding share by a few points. The other thirty-something points would still be there.

The intro is a better guess, and it is still not the answer. The median video starts paying off at 1:00 — which means the setup people complain most about is usually a minute, not five. What actually costs you is distributed: the recap after each section, the restatement of the thesis in slightly different words, the anecdote that illustrates a point already made, the closing summary of a video you have just finished watching.

None of that is deceptive. Each individual instance is defensible. It is the accumulation that turns fifteen minutes of content into a forty-minute video.

Longer videos are less padded, not more

Here is the result that surprised us most, and it runs against the entire folk theory of YouTube.

We expected padding to rise with runtime — that the long videos would be the bloated ones, stretched to hit some threshold. The measurement says the opposite. Videos under ten minutes are the most padded thing we measure. Padding share falls steadily as runtime grows, and the videos over an hour are the least padded material in the library by a clear margin.

Once stated, the reason is not mysterious. A subject that genuinely needs an hour fills an hour. A subject that needs six minutes and is being delivered in eleven has to find five minutes somewhere, and the only available material is restatement.

The practical version: runtime is close to useless as a signal of whether a video will waste your time. A nine-minute video is not a safe bet, and a fifty-minute one is not automatically self-indulgent.

Which subjects stall most

The spread across categories is wider than the spread across lengths, and it lines up with how compressible the subject is.

The most padded categories we measure are design, productivity and language learning — all three built around ideas that can be stated in a sentence and therefore have the furthest to stretch. A productivity video’s core method is frequently one line. The remaining nineteen minutes are structural.

The least padded are science and history. Both deal in material where the next minute depends on the previous one: a causal chain does not compress into a list without becoming a different, worse video. History has the second-longest average runtime of anything we measure and the second-lowest padding share, which is exactly the combination the folk theory says should not exist.

What the title promises

There is a fourth measurement worth mentioning here, because it correlates with padding more tightly than anything else: whether a video’s title describes what is inside it.

Videos whose titles are accurate have a padding share well below the library average. Videos whose titles substantially oversell run around half again as high, and in the small group where the title is essentially a different video from the contents, padding runs past 70%.

That relationship makes sense in one direction and is worth stating carefully. A title written to attract a click rather than to describe the contents belongs to a video optimised for the same thing, and a video optimised for the click has less reason to respect the runtime. It is not a rule — plenty of honestly titled videos ramble — but as a prior it holds up across the library.

What to actually do about it

Three things, in descending order of how much time they save.

Stop treating length as the signal. It is not one. Judge the density instead, and if you cannot judge it from outside, that is the whole problem this measurement exists to solve.

Start where the video starts. For a video that takes 1:00 to arrive, that is a minute. For the one in ten that takes 3:46 or longer, it is most of the reason you were about to abandon it.

Know the shape of the subject. In a category built on one compressible idea, the fifth video adds almost nothing to the first four — the marginal return has gone, and the density measurement shows it going. In a subject that builds, the long one is frequently the honest one.

None of this requires a tool. It requires knowing the ratio before you commit an hour, which — until you have watched the video — is the one thing the page does not tell you.