What Is a Good Retention Rate for YouTube Shorts?

YouTube publishes no official retention benchmark for Shorts. Its own documentation defines the metrics — average percentage viewed, engaged views, how many chose to view — and states no target figure for any of them. The number worth measuring against is your own channel's median, taken across clips of similar length read at the same age.

Does YouTube publish an official retention benchmark for Shorts?

No. YouTube documents the metrics with some care and then declines to score them.

The YouTube Analytics API reference defines averageViewPercentage as "The average percentage of a video watched during a video playback." Inside Studio, YouTube's help page on content performance describes the same metric as "Average percentage of a video watched among those who stayed to watch." Both sentences explain the arithmetic. Neither names a number to beat.

The closest YouTube comes to a verdict is a metric called relativeRetentionPerformance, defined in that same API reference as "A measurement that shows how well a video retains viewers during playbacks in comparison to all YouTube videos of similar length."

Read that definition twice. YouTube's own answer to "is this good" is a comparison, not a constant — and it is bracketed by length. That is the whole method, handed to you by the platform.

On the Shorts view of the Content tab, YouTube lists "Shown in feed" — "The number of times that your Shorts showed in the Shorts Feed" — alongside "How many chose to view", which "Highlights the percentage of times that viewers viewed your Shorts versus swiped away." No target sits beside either one.

Why do the published Shorts retention benchmarks contradict each other?

Because they are separate estimates from separate sample sets, and they disagree — often about the same quantity.

Pick an arbitrary reading. Say a 45-second Short at 45% average percentage viewed. Here is that one number, judged by four published scales:

Source	What it publishes	Verdict on 45%
Retensis	"On Shorts between 30 and 60 seconds, aim for above 40%."	Above target
Lenostube	Shorts average: "Retention Rate: 30% – 50%"	Perfectly average
Prepublish	"If your retention sits below 70 percent, your hook or pacing is likely the problem."	Broken hook
Humble&Brag	"Below 50 per cent indicates a structural problem"	Structural problem

Same clip, same number, and depending on whose post you opened first you either beat the target or need to rebuild the video.

The disagreement is not confined to targets. On what Shorts average, Lenostube's table says "30% – 50%". Virvid reports "YouTube Shorts average ~73% viewer retention according to Socialinsider". Those two sentences describe the same population and sit more than twenty points apart. At least one of them is wrong, and reading both will not tell you which.

Credit where it is owed: Retensis states plainly that its figures are "targets, not measured platform averages" and that "Nothing on this page claims to describe the average video on any platform." That is the honest way to publish a number. Several of the others present theirs as settled fact.

Which retention numbers can you actually see on a Short?

Fewer than the benchmark posts assume.

YouTube's audience retention report — the curve that shows exactly where viewers leave — carries a stated requirement on the help page that documents it: "Your video should also be at least 60 seconds long and have at least 100 views." Shorts run up to 3 minutes, so a long one clears the bar. A 30-second Short does not. Humble&Brag's advice to "hold above 80 per cent in the first three seconds" is prescribing a shape read off a graph that, for any Short under a minute, is never drawn at all.

What every Short does give you is the per-clip summary: views, engaged views, average percentage viewed, average view duration, and the feed pair above. For where those cards sit in Studio and what each one counts, we covered that separately in how to read YouTube Shorts analytics.

How do you build a retention baseline from your own channel?

Five steps, no borrowed numbers.

1. Fix the age. Read every clip at the same age — 48 hours, or 7 days. A six-month-old Short has collected a long tail of feed traffic a two-day-old one has not. Mixing ages is the most common way a homemade baseline goes wrong.
2. Bracket by length. Group 0–20s, 20–45s, 45–90s, 90s+, or whatever bands match your output. YouTube's own comparative metric is defined against videos "of similar length"; a 15-second punchline and a 2-minute story are not the same measurement.
3. Collect at least ten. YouTube's retention report uses "typical retention to compare your 10 latest videos of similar length." Ten is the platform's own choice of sample floor, which makes it a defensible one for yours.
4. Take the median, not the average. Sort the percentages low to high and take the middle value; with an even count, average the middle two. One runaway clip drags a mean upward and quietly raises a bar nothing else in the set can clear. The median ignores it.
5. Recut it quarterly, and after any real format change. A baseline built around talking-head clips says nothing useful about the month you switched to on-screen text.

Fixing the age in step one is the fiddly part, because Studio always shows you now. Everpop's signed 48h and 7d YouTube Analytics receipts were built for this: each is a link a third party can open showing what a published clip did at those two fixed ages, flops included. A receipt documents an outcome — it never predicts one — and fixed ages are precisely what makes one clip comparable to the next.

Once the baseline exists, read it as a distribution rather than a line. A single clip under the median is noise; four in a row under it is a signal worth acting on.

Does the August 2026 view-counting change affect your baseline?

It is a good reason to date-stamp every benchmark you read.

Per YouTube's page on how engagement metrics are counted: "Beginning August 24, 2026, views are counted the moment a video starts to play across all formats, including Shorts, long-form videos (VOD), and live streams." The money side did not move with it — "YPP earnings will still be based on 'engaged views' and 'engaged watch hours'."

Counting rules shift. A "2026 benchmark" written before a counting change is describing a slightly different measurement than the one your dashboard now performs, and no vendor post updates itself. Your own median does, the moment you recut it. If the views-versus-engaged-views split is new to you, start with what counts as a view on YouTube Shorts.

What should you do when a Short lands under your median?

Treat it as one observation, then look for the pattern.

Check the length band first — a clip that fell short of the 45–90s median may be sitting comfortably above the 0–20s one, and the fix is a shorter cut rather than a better hook. Check the feed pair second: a strong "how many chose to view" with weak average percentage viewed points at the middle of the clip, while a weak one points at the frame and the first second.

Nobody's median is a grade. It is a control line that tells you which change is worth testing next — and unlike a number from a vendor blog, it is measured on your audience, your subject matter, and your cuts.