Does a longer video teach you more? We checked, and the answer is yes

We have already published the result that runtime is a poor filter, because longer videos carry less padding rather than more. That answers a question about waste. It does not answer the question people actually ask, which is whether the extra half hour contains anything.

It does. Average measured depth rises from 48.3 for videos under ten minutes to 79.7 for videos over an hour — thirty-one points on a hundred-point scale, and the single largest relationship between runtime and any measurement we publish.

A note on the numbers. Unlike the rest of this site, the figures on this page are frozen at the date above rather than recalculated on each visit. A report people cite should not move under them. When we run this measurement again it will be published as a new edition with a new date.

Depth climbs, and it climbs at every step

Depth is one of the readings behind the headline value score: how far past the surface of its subject a video goes, as against how much of it is definition and framing.

Every step up in runtime is a step up in depth, with no reversal anywhere in the sequence. The largest single jump is the first — 11.2 points from the sub-ten-minute band to the ten-to-twenty band — which is worth noticing, because it means the interesting boundary is not “long video” versus “short video” but “under ten minutes” versus everything else.

Novelty moves with it, from 34.7 to 72.5. That is the reading for how much of a video is material a viewer is unlikely to have met already, and it more than doubles across the range.

It holds inside a subject, which is the check that matters

The obvious objection to any result like this one is composition. If the long videos in the library are mostly science lectures and the short ones are mostly productivity tips, then the chart above is measuring the difference between science and productivity, not the difference between long and short.

So we ran the same measurement inside single subjects.

Depth rises with runtime in all six, and in four of them it rises at every step. Science goes from 56.5 to 85.0, business from 43.5 to 64.2, programming from 50.2 to 67.3. The two reversals are both in the thirty-to-sixty-minute band of a subject where that band holds fewer than fifty videos.

That is not proof of a mechanism, but it removes the easy alternative explanation. Within a subject, holding the topic roughly constant, the longer video goes further.

What length does not buy

Read the full table and one column refuses to move with the others.

Applicability — whether you could act on what the video told you — goes from 51.5 to 55.3 across the first four bands while depth goes from 48.3 to 67.8. Only in the hour-plus band, which holds fifty videos, does it climb sharply. A longer video reliably goes deeper; it does not reliably become more useful.

That distinction is the practical content of this report. If you want to be able to do something by this evening, the extra forty minutes is buying you very little on the measurement that matters to you. If you want to understand why the thing works, the extra forty minutes is buying you almost everything.

The level mix says the same in a different form. Videos under ten minutes are 70.6% introductory; videos over thirty minutes are around 29%. A short video is not a worse video, it is an introductory video, and there is nothing wrong with that unless you already have the introduction.

The skip rate falls, and one band is empty

30.2% of videos under ten minutes end up with a skip verdict, against 12.5% between thirty and sixty minutes. In the hour-plus band there are no skips at all.

The zero is not a finding, and we would rather say so than let it stand as one. Fifty videos is a small band, and a library assembled for a recommendation product is going to contain hour-long videos that somebody thought were worth the hour. That is selection, not evidence about the platform. Read the trend across the first four bands, which rest on 3,193 videos between them, and treat the last bar as an illustration.

How this fits with the padding result

Both results point the same way once you stop treating runtime as a cost.

Padding falls with length because a subject that genuinely takes an hour fills an hour, while a subject that takes six minutes and is delivered in eleven has to find five minutes somewhere. Depth rises with length for the same reason from the other side: the videos that run long are, in this library, the ones with something that takes that long to say.

The common cause is the subject, not the runtime. Length is a symptom of a topic with room in it — which is why length works as a weak signal of depth and fails completely as a filter for waste.

What this measurement does not say

It is not causal. Nothing here says that making a video longer would make it deeper. The reasonable reading is that both length and depth follow from what the video is about and how seriously it was made.

Depth is not quality. A deep video can be badly explained, and a shallow one can be exactly what somebody needed. Depth measures how far past the surface the material goes, and that is useful precisely when you specify what you want it for.

It is not accuracy. We measure how much a video says and how far it goes, not whether what it says is true. A confidently argued, deeply developed video about a claim that does not hold up scores well here.

How this was measured, and what it does not say

Depth, novelty and applicability are three of the readings behind the value score; every one is on a 0–100 scale and all three are ours. Method and its limits are in what value density means.

The readings do not all cover the same set. Depth, novelty and applicability are read on 3,137 of the 3,243 measured videos; padding on 3,002, because it is only measured on explanatory and instructional material. Each figure states the set it comes from.

The hour-plus band holds fifty videos. Every number in that row is an illustration rather than an estimate, and the report’s argument does not depend on it — the trend is already unambiguous across the four bands below.

The sample is not YouTube. It is English-language explanatory video across twenty subjects, assembled to build a recommendation library rather than to be representative of the platform. Long videos that nobody thought worth measuring are, by construction, absent.

The figures are frozen at 1 September 2026. Scores are recalculated as the underlying measurement improves, so the live numbers on the shelves will already have moved slightly from these. This edition will not be edited; the next measurement will be published separately.