Published 29 September 2026 · Figures measured 15 September 2026 · Every figure below carries the number of Shorts behind it
What the bars in circulation say
Search for a retention benchmark for Shorts and you get a remarkably consistent answer. We read these pages on 29 September 2026 and this is what they say, quoted:
- humbleandbrag.com, "YouTube Shorts Benchmarks 2026": "A healthy average percentage viewed for Shorts in 2026 is 70 per cent or above." Below 50% "indicates a structural problem." No per-length bands. No sample size is given for either figure.
- virvid.ai, "The Best Length for Shorts (Retention Data Breakdown)": a per-length table giving 73 to 85% under 20 seconds, 65 to 75% at 20 to 40 seconds, 60 to 76% at 40 to 60 seconds and 25 to 50% from one to three minutes. It discloses no dataset of its own and cites six other blogs.
- shortimize.com, "YouTube Shorts Retention Rate (2026)", which is the most careful of them: "A 12-second Short and a 2:45 Short shouldn't have the same APV expectation." It gives no numbers for that, and no sample size.
Two things are worth noticing before any of our own data arrives. The first is that the shape of the advice is right: expectation really should fall with length, and the third page says so without a table behind it. The second is that not one of these pages states how many Shorts it looked at, and several of the numbers are cited from one another rather than measured.
So we checked both halves separately: does the expectation actually fall with length, and does clearing a threshold actually predict anything. They have different answers.
What our catalogue reads, by length
Every Short in the table below is a real Short on a real customer channel, measured on 15 September 2026. The base is 575 Shorts across 18 channels holding measurement data.
| Length | Shorts | Median % viewed | p25 | p75 | Median views |
|---|---|---|---|---|---|
| Under 20s | 325 | 109.7% | 85.7% | 149.6% | 131 |
| 20 to 40s | 92 | 58.8% | 47.9% | 81.0% | 478 |
| 40 to 60s | 53 | 59.0% | 42.9% | 73.9% | 98 |
| Over 60s | 105 | 44.0% | 25.8% | 60.5% | 72 |
Three observations, in the order they matter.
The under-20-second band sits above 100%. That is not a band of unusually gripping videos. A Short restarts when it ends, so a 10 second Short watched once and a bit reads over 100%, and a band short enough for that to happen easily will have a median above the line. What a number above 100% means, how common it is and what it predicted are a separate question with a separate measurement behind it, and it has its own page here. For this page the only thing that matters about it is that 109.7% is not a target, and a 55 second Short that never approaches it is not failing at anything.
The middle two bands are indistinguishable here. 58.8% and 59.0% is not a difference, on 92 and 53 Shorts. Pooled across channels, this table cannot tell 20 to 40 seconds apart from 40 to 60. That is a real limit of the pooled view, and it is the reason the next table exists.
And the pooled table is weak evidence anyway. The bands are not the same channels. If short Shorts happen to be made by one creator and long ones by another, a length effect and a channel effect look identical. So the question has to be asked inside a single channel.
Inside one channel, it falls every time
This is the pass that decides it. For each channel with at least four Shorts in two or more length bands, here is that channel's own median, band by band. Comparing bands within one channel removes topic, audience and channel size at a stroke, because it is the same creator either side of the comparison.
| Channel | Under 20s | 20 to 40s | 40 to 60s | Over 60s |
|---|---|---|---|---|
| A | 214% (n=15) | 115% (n=7) | · | · |
| B | 108% (n=151) | 51% (n=17) | 41% (n=9) | 28% (n=6) |
| C | · | 48% (n=11) | · | 33% (n=4) |
| D | 105% (n=56) | 63% (n=10) | · | 36% (n=6) |
| E | 83% (n=25) | 62% (n=21) | 58% (n=18) | 55% (n=50) |
| F | 82% (n=6) | 48% (n=6) | · | · |
| G | · | 76% (n=14) | 69% (n=10) | · |
Take every pair of neighbouring bands a channel actually has data for. There are 12 of them across those 7 channels, and in all 12 the shorter band has the higher median. Not eleven. Twelve.
That is the strongest single result we hold on this subject, and it is the one thing on this page we would defend without hedging. A 45 second Short at 59% viewed and a 10 second Short at 59% viewed are not the same object, and comparing either against a single number is comparing them against the wrong thing.
It is also worth being clear about part of the mechanism, because it is arithmetic rather than craft. Some of the fall comes from the fact that a 9 second Short can be watched twice inside the time it takes to lose interest and a 61 second one cannot. That does not weaken the finding for the purpose it is used for here, but it does mean the gap between the shortest band and the longest is partly a property of the format and not only of how well anybody held an audience.
Now the bar. We tested it, and it did not predict
Knowing that expectation falls with length is not the same as having a threshold. So we took a per-length threshold table that was put in front of us and asked whether clearing it meant anything. Here are the exact bars we tested, printed in full so anyone can repeat the test or disagree with it:
Then the outcome. We defined success as the Short reaching 1,000 views, and separately 4,000, and asked the question both ways: of the Shorts that cleared the bar, how many got there, and of the Shorts that missed it, how many got there.
| Length | Shorts | Base rate, reached 1,000 | Cleared the bar | Of those, reached 1,000 | Missed the bar | Of those, reached 1,000 |
|---|---|---|---|---|---|---|
| Under 20s | 325 | 28.6% | 231 | 34.2% | 94 | 14.9% |
| 20 to 40s | 92 | 38.0% | 24 | 37.5% | 68 | 38.2% |
| 40 to 60s | 53 | 28.3% | 21 | 28.6% | 32 | 28.1% |
| Over 60s | 105 | 4.8% | 15 | 0.0% | 90 | 5.6% |
In the 20 to 40 second band, the Shorts that cleared the bar reached 1,000 views 37.5% of the time and the Shorts that missed it reached 1,000 views 38.2% of the time. In the 40 to 60 second band it is 28.6% against 28.1%. Those are the same number. In the over 60 second band it points the wrong way: none of the 15 Shorts that cleared the bar reached 1,000 views, and 5.6% of the 90 that missed it did.
Only the under-20-second band separates at all, 34.2% against 14.9%. That band is also the one where the bar sits at 90% while the band's own median is 109.7%, so most of the band clears it: 231 of its 325 Shorts are on the clearing side. What a reading above 100% actually means is a separate measurement and belongs to the other page, not this one.
Half the catalogue clears it
Pooled across all four bands, 291 of our 575 Shorts (51%) cleared the "viral" threshold for their own length. That is the finding, and everything else on this page is context for it. A bar that half a catalogue of small-channel Shorts clears is not identifying anything unusual.
Of those 291 Shorts:
- 32.3% reached 1,000 views. Two in three did not.
- 3.8% reached 4,000 views. 96.2% never did.
- 1.0% reached 10,000 views. 99% never did.
The median Short that cleared the bar took 191 views. The median Short that missed it took 96. There is a difference there and we are not going to pretend otherwise, but a creator told they had hit a viral threshold and then handed 191 views has not been told anything they can use.
And the level does not transfer between channels either
Even if a bar of some kind were useful, it could not be one number, because the same length band reads completely differently on different channels. Here is the median under-20-second Short on every channel in the base with at least four of them.
| Channel | Shorts under 20s | Median % viewed |
|---|---|---|
| Channel 1 | 15 | 213.7% |
| Channel 2 | 53 | 129.8% |
| Channel 3 | 151 | 107.7% |
| Channel 4 | 56 | 105.1% |
| Channel 5 | 7 | 94.3% |
| Channel 6 | 25 | 82.5% |
| Channel 7 | 6 | 81.9% |
Top to bottom that is a 2.6x spread on one length band. Channel 7's median Short sits below the bar we tested and Channel 1's median Short clears it by more than a hundred points.
Channel 1's best Short, of any length, has 1,855 views.
So what should you compare your number against?
Your own catalogue, in your own length band. That is not a hedge, it is the only reference point our data supports, and you can build it yourself in about ten minutes with no tool at all.
- List your Shorts with their length and their average percentage viewed. YouTube Studio gives you both on the Content tab.
- Sort them into length bands. Under 20 seconds, 20 to 40, 40 to 60, over 60 is what we used. Use whatever splits leave you at least four Shorts in a band. We used a floor of four, and below that the middle value moves around too much to lean on.
- Take the middle value in each band. That is your channel's own expectation for that length.
- Read a new Short against its own band and nothing else.
What that buys you. It stops a long Short being judged against a short one, and on our data that mistake is large. Channel B in the table above reads 108% at under 20 seconds (n = 151) and 28% over a minute (n = 6), inside one channel, on one audience. An 80 point gap is not a quality difference, and a bar that ignores it will call that creator's long Shorts a disaster every time.
What it does not buy you, and we would rather say it than let you find out. A band table will not tell you which of two Shorts in the same band will do better. We tested exactly that on the same 575 Shorts: comparing Shorts from one channel inside one length band, the Short with the lower average percentage viewed had more views 44.6% of the time. That is close enough to a coin flip that average percentage viewed, on its own, does not rank a creator's own Shorts against each other.
So the honest description of what a band table is for: it tells you whether a Short landed far below what your channel normally does at that length. It does not tell you which of the ones that landed normally will travel. Why that is, and what it means for reading any retention number, is a wider measurement than this page and is not about length at all.
The limits, before you use any of this
These are small channels. The band medians in the first table run from 72 views to 478, and roughly one Short in a hundred anywhere in this catalogue has ever passed 10,000 views. Nothing here describes a channel that routinely reaches six figures and it should not be read as if it does. A bar calibrated on large channels is not wrong for being different from ours. It is wrong for being quoted at people like ours.
Two channels supply a lot of this. The under-20-second band is 46% one channel and the over-60-second band is 48% another. That is exactly why the within-channel table exists and why the 12 of 12 result is the one we lean on: it cannot be produced by a channel mix, because every comparison in it happens inside one channel.
Retention is measured after distribution, not before it. A Short that gets shown to more people is shown to better-matched people, and its measured retention rises with that. So any reading of "high retention went with more views" is flattered by the order the two things happen in. That bias favours the bar we tested, which makes it more striking that the bar still separated nothing in three bands out of four.
One band is thin. The over-60-second result rests on 15 Shorts clearing the bar. 0.0% of 15 is a real measurement and it is also a small one, and we would not build anything on that row alone. The 20 to 40 and 40 to 60 second rows, at 92 and 53 Shorts, are what carry the conclusion.
What this page does not say
- Not "retention does not matter." What we tested is narrow: whether a published threshold, applied inside a length band, separated the outcomes of the Shorts inside that band. It did not, in three bands out of four. That is a different question from whether retention is related to views at all, and this page did not set out to measure that one.
- Not "our medians are the real benchmark." Publishing our table as the global numbers would be the identical mistake to the one this page is about, one catalogue further along. The direction transfers because it held inside every channel we could test it in. The levels are ours.
- Not "make your Shorts shorter." We measured different Shorts at different lengths. We did not measure the same Short cut two ways, and nothing here tells you what would happen if you cut yours.
- Nothing about what YouTube does with any of these numbers. We cannot observe it and we are not going to guess. Every sentence on this page is about Shorts, lengths and view counts we can see.
And the thing we genuinely cannot see
The metric most accounts of Shorts distribution actually point at is viewed versus swiped away, the share of people who were shown the Short and stayed rather than flicking past. It is visible in YouTube Studio and it is not in the data a creator can authorise us to read. Every threshold in circulation for it is therefore untestable here, and average percentage viewed is not a proxy for it: it describes the people who started watching and says nothing about how many never did. We will not publish an estimate of it and we will not tell you whether you clear anybody's bar for it. We explain what happened to the people who started. Studio tells you how many never did.
Questions people ask about this number
Taken from search results in September 2026, close to the wording they were asked in. Every answer carries the number of Shorts behind it, and where the honest answer is that we do not know, that is the answer.
Will cutting my Short in half raise my average percentage viewed?
Mechanically, yes, if viewers leave at the same moment they were leaving before. The figure is time watched divided by the length of the Short, so making the length smaller makes the figure bigger. Whether they would leave at the same moment is the part we cannot tell you, and it is the part that decides whether anything real improved. We measured different Shorts at different lengths. We have never measured the same Short cut two ways, and no page we found has either.
What we can tell you is that the expectation genuinely is lower for a longer Short, so a longer Short is not underperforming merely by reading lower. Across 7 channels with data in two or more length bands, all 12 neighbouring band comparisons ran the same way, with the shorter band higher, measured 15 September 2026.
And it is worth asking what the higher number would be for. On the same base of 575 Shorts, clearing a published threshold for a band and missing it gave the same rate of reaching 1,000 views in the 20 to 40 second band (37.5% against 38.2%, n = 92) and in the 40 to 60 second band (28.6% against 28.1%, n = 53). Raising a ratio that did not separate those outcomes is not obviously the thing to spend an edit on.
Is 70% a good average view percentage for a Short?
It is not a complete question until you say how long the Short is, and that is the most useful thing we can say about it.
On 575 Shorts across 18 channels, measured 15 September 2026, 70% would sit well below the middle of our under-20-second band (median 109.7%, n = 325), a little above the middle of both middle bands (58.8% at n = 92 and 59.0% at n = 53), and comfortably above the middle of our over-a-minute band (44.0%, n = 105). The same 70% is three different verdicts depending only on the length of the video.
It also moves by channel, not just by length. Among channels in that base with at least four Shorts under 20 seconds, the median ran from 213.7% (n = 15) down to 81.9% (n = 6), a 2.6x spread on one band. A single number cannot be right for both of them.
My 60 second Short only holds 40%. Is something wrong with it?
On our data, 40% on a Short over a minute is close to ordinary. The median Short over 60 seconds in our base read 44.0% viewed (n = 105, measured 15 September 2026), with the quarter-point at 25.8% and the three-quarter point at 60.5%. A 40% reading sits inside the middle half of that band.
The advice in circulation says otherwise, and it is worth naming the gap rather than talking around it. One of the pages we read says plainly that below 50% "indicates a structural problem", and gives no length band and no sample size for it. On our 105 Shorts over a minute, that sentence would call a structural problem on more than half the band, the median Short included, because that band's median is 44.0%.
It would also be calling it on the wrong side. In that band, the Shorts that cleared the threshold we tested reached 1,000 views less often than the ones that missed it: 0.0% of the 15 that cleared against 5.6% of the 90 that did not.
What we would actually look at is whether 40% is low for your channel at that length, which is a table you can build yourself in ten minutes. There is a walkthrough above.
Is there a retention number that gets a Short pushed to more people?
We cannot answer that, because it asks what YouTube does, and we have no view into that and will not guess at it.
What we can report is whether Shorts that cleared a published threshold went on to do better than Shorts that did not, on a catalogue we neither designed nor influenced. On 575 Shorts across 18 channels, measured 15 September 2026, 291 of them (51%) cleared the per-length "viral" threshold we tested. Of those 291, 3.8% reached 4,000 views, so 96.2% did not. The median Short that cleared the bar took 191 views.
And within three of the four length bands, clearing the bar and missing it gave the same outcome rate, or a worse one. The table is above. That is a record of what happened after the fact. It is not a statement about a mechanism, and it does not tell you that a threshold plays no part in anything. It tells you that this threshold, on this catalogue, sorted nothing.
Two questions this page deliberately does not answer. What a reading above 100% actually means, and what it predicted, has its own measurement and its own page: is over 100% retention good on a YouTube Short? Why a Short with better retention lost to one with worse retention is a wider measurement than anything here and is not about length at all.
How this was measured
Why we looked at this at all
ShortLens compares each of your Shorts against your own recent Shorts rather than against a published benchmark, and reads the measured retention curve to tell you where viewers left, to the second. This analysis is a large part of the reason it works that way. A product that judged a 70 second Short against a 70% bar would be judging it against a number that separated nothing on our own over-a-minute Shorts: of the 15 that cleared the threshold we tested, none reached 1,000 views, against 5.6% of the 90 that missed it.