Does YouTube Shorts algorithm retention decide whether a video keeps running?

What to take away

• YouTube Shorts algorithm retention is read as average percentage viewed, not total minutes watched.
• The first three seconds carry most of the loss, and a strong back half rarely repairs a weak opening.
• Pick one numeric line before you open Studio, then judge on a fixed date.
• Retention says nothing about who watched or what they did next.
• Payouts track reach, so a retention slide reaches your revenue weeks later.

The metric worth tracking

Average percentage viewed is the share of a Short's running time that the average viewer watched, expressed as a percent. A 40 second Short that the average viewer leaves at second 22 scores 55 percent. YouTube Studio reports that figure beside a second split, watched against swiped away.

Shorts sit in their own feed with their own signals, as the Wikipedia entry on YouTube Shorts sets out. Total watch time hides the shape that matters. A 60 second Short can collect more minutes than a 15 second clip while holding a smaller share of its audience.

Length bands decide how the metric reads. A feed algorithms comparison in plain terms keeps Reels, Shorts and the TikTok feed from blurring into one set of assumptions.

How to read a retention curve before you act

1. Open the Short in YouTube Studio and set the range to the first 14 days it was live.
2. Read the watched and swiped away split first, because it caps everything underneath.
3. Find the second where the curve first flattens and write it down.
4. Compare average percentage viewed with your trailing median for the same length band.
5. Compare against Shorts from the same week before comparing across months.

YouTube publishes no pass mark, so you set one. A workable starting line: act when a Short sits more than 15 points below the trailing median for its length band and the same format does that three times running. Those figures are illustrative, not official.

What retention cannot tell you

Small view counts make the number swing. A Short with a few hundred views can move ten points in a day. Wait until the curve stops shifting between refreshes.

The metric is length dependent, so a 15 second Short and a 55 second Short cannot share one chart. Loops and replays land in the same bucket as first watches.

It also cannot separate a dull Short from a good one shown to the wrong audience. Sifting real movement from weekly noise is the harder skill, and tracking platform trends without the noise saves more time than another dashboard.

Example: a 38 second Short that lost the room at second six

A creator posted a 38 second Short that scored 34 percent average percentage viewed and 61 percent swiped away. The curve dropped hard to second six, then ran nearly flat to the end. The hook, a question, arrived at second nine.

Viewers who stayed stayed. The ones who left never heard the setup. The next post moved that line to second one and cut seven seconds of wind-up. That is a diagnosis, not proof of cause.

Attribution and its limits

Posting hour, audio choice and topic all move alongside retention. So does the mix of people the Short reached. A rise after a change is not proof the change worked.

YouTube does not publish the weights behind the feed, so ranking is a system you test rather than read. The evaluation material on NIST's AI topic page covers that discipline for a recommendation feed as much as for a model.

Keep the comparison inside one length band, one topic and one week. Metric names in Studio shift without notice, so feature updates research is worth reading when a figure you relied on disappears.

When to stop measuring and decide

Set the window before the batch goes out: 14 days per Short, or 10 Shorts per format, whichever comes first. Stop when the trailing median holds within a couple of points, week over week. Then decide, and do not reopen the question mid batch.

• The batch has run its full window.
• The format median has been flat for two weeks.
• The same format has cleared or missed the line three times.

Keep the format if it clears the line across the batch. Retire it if it misses. Anyone who has learned to read an annual trend report will spot the difference between a weak format and a weak quarter.

Payout rules sit on the YouTube Help page on Shorts monetization, which ties earnings to views rather than retention. A retention slide reaches your balance weeks after the curve turns.

Common questions

Does retention decide whether a Short keeps running?

It is the clearest signal you can see, and not the only one. YouTube weighs signals it does not publish, so treat retention as necessary rather than sufficient.

How long before a Short's numbers mean anything?

About 14 days for the curve to settle, and longer for a channel median. A Short still climbing on day three has not finished.

Do payouts follow retention directly?

No. The chain runs from retention to reach to views to payout rate, and the rate moves month to month.

Can one Short settle a weak format?

No. A single result is noise. Wait for three misses in a row before retiring anything.