How long should a YouTube video be?

Short answer: as long as the idea needs, and not a minute longer. The data backs this up more cleanly than any guru's rule ever has.

Every creator has heard a number. Eight minutes, because of mid-roll ads. Ten minutes, because that used to be the magic threshold. Fifteen, because the algorithm loves watch time. Twenty, because deep content wins. The numbers contradict each other, they change every year, and they all miss the point in the same way: length is not a target. It is a consequence — of how much you have to say, and how well you say it.

What the retention data actually says

Let us start with what is measured, not what is preached. Across YouTube, the average video holds about 24% of its viewers to the end. More than half leave within the first minute. Those are platform-wide averages, dragged down by everything from abandoned uploads to clickbait — but the pattern by length is consistent everywhere you look:

• Under 5 minutes: 50 to 70% retention is healthy.
• 5 to 10 minutes: 40 to 55%.
• 10 to 20 minutes: 35 to 45%.
• Over 20 minutes: 25 to 35%.

The rule is simple and unforgiving: the longer the video, the lower the percentage. Every extra minute is another chance for someone to leave. A 30-minute video holding 35% is doing genuinely well. A 4-minute video holding 35% has a problem.

But here is the part people miss: percentage is not the score. Watch time is.

Percentage lies, watch time tells the truth

Do the math on two videos. A 4-minute video with 70% retention delivers 2.8 minutes of average watch time. A 30-minute video with 35% retention delivers 10.5 minutes. The second video "retains" half as well by percentage and generates nearly four times the watch time — and watch time is what the recommendation system actually weighs.

This is why the "shorter is better" advice is half-true. Shorter videos retain a higher percentage, which looks better on the graph and feels better to the creator. But the algorithm does not feel. It counts minutes. A long video that holds a third of its audience is often doing more work for a channel than a short video that holds most of it.

None of this means long is automatically better. A 30-minute video that holds 15% is 4.5 minutes of watch time wrapped in 25 minutes of people leaving. Length amplifies whatever the content already is. Good pacing gets more room to compound. Bad pacing gets more room to bleed.

The real question: how long does the idea need?

Forget the numbers for a moment and ask the question the numbers are trying to answer. Every video makes a promise — a question it will answer, a thing it will show, a story it will tell. The right length is however long it takes to deliver on that promise completely, at a pace that respects the viewer's time.

Some ideas are four minutes long. "Here is the setting that fixes it" does not need a backstory. Stretching it to ten minutes does not add value; it adds the exact footage where your retention graph falls off a cliff. Viewers can feel padding the way you can feel a mattress lump. They leave, and they are right to.

Some ideas are twenty-five minutes long. A real investigation, a genuine build, a story with actual stakes — these need room. Cutting them to eight minutes does not make them punchier; it makes them shallow, and shallow does not get shared.

The creators who obsess over length are usually avoiding a harder question: is this video dense enough for its runtime? Density — value per minute — is the actual variable. A dense 20-minute video beats a thin 8-minute video every time, on every metric that matters.

Sweet spots by format

That said, formats have shapes, and the shapes exist for reasons. These are starting points, not laws:

• Tutorials and how-tos do well at 5 to 12 minutes. The viewer arrived with a question; answer it, show the steps, and stop. Educational content averages some of the highest retention on the platform — around 42% — because the intent is strong. Do not dilute strong intent with weak filler.
• Video essays and commentary live comfortably at 12 to 25 minutes. The audience opted into thinking, and thinking takes time. The risk is not length but drift — one section too many, and the graph sags in the middle.
• Vlogs and entertainment are the hardest: 8 to 15 minutes, with ruthless cutting. Casual viewers make constant stay-or-leave decisions, and vlog retention averages around 21% — the lowest of any major category. Every boring minute is a referendum you lose.
• Reviews and comparisons work at 8 to 15 minutes. Long enough to be genuinely useful, short enough that the viewer does not just skip to the verdict — though many will, which is fine.
• Podcasts and interviews can run 30 minutes to over an hour, because the consumption mode is different. People listen while doing other things. Retention percentage will look terrible and it does not matter; the absolute watch time is enormous.

Notice the pattern: the length follows the viewer's intent, not the creator's ambition.

The padding trap

There is one specific way creators get length wrong, and it deserves its own section because it is so common: padding for ad breaks.

The old logic went like this — longer video, more mid-roll ads, more revenue. So creators stretched 6 minutes of content into 12. Intros got longer. Recaps appeared mid-video. "But wait, there's more" became a structural device.

The problem is that padding does not just bore viewers; it trains the algorithm against you. Every padded minute lowers retention, and lower retention means fewer recommendations, which means fewer views, which means the extra ad slots play to an audience that already left. It is the rare strategy that fails on both sides of the ledger: less reach and less revenue per viewer, because the viewers who stay are the least engaged ones.

YouTube has also quietly reduced the incentive. Mid-roll eligibility and ad load are no longer the simple "8 minutes unlocks everything" game they once were. The platform's own guidance has shifted toward satisfaction — did the viewer feel their time was well spent? A tight 7 minutes that satisfies beats a padded 14 that does not, in recommendations and in revenue.

The exceptions that prove the rule

Every guideline above has exceptions, and the exceptions are instructive.

Some channels break the retention math entirely. True-crime narration, video essays with devoted followings, creators whose audience treats new uploads like appointments — these can hold 50% or more on 25-minute videos, numbers that look impossible on the benchmark tables. What they share is not a trick. It is trust, built over years: the audience arrived already believing the time would be well spent, so they grant the video a longer runway.

Live streams play by different rules too. A two-hour stream with 15% retention sounds catastrophic until you do the arithmetic: 15% of two hours is eighteen minutes of watch time per viewer — more than most edited videos deliver in total.

And then there is the small-channel effect. Loyal audiences retain dramatically better than large casual ones, because every viewer chose to be there. A 3,000-subscriber channel can hold 60% on a 12-minute video while a 3-million-subscriber channel holds 35% on the same length. The algorithm knows this, which is why it compares you against similar videos rather than against the biggest channels on earth.

The lesson is not that benchmarks are useless. It is that they describe the average, and you are not obligated to be average. Build the trust, earn the runway, and the numbers bend.

How to find your length

So how do you actually decide, video by video? You already have the instrument. It is in YouTube Studio, under every video you have ever published.

Open your retention graph and look at the shape, not just the number. A steady decline means the length is roughly right and the pacing is even. A cliff in the first minute means the hook failed, not the length. A sag in the middle means a section overstayed — cut it next time, or cut it now. Spikes where people rewatch tell you what they actually came for; consider making that the whole video next time.

Then compare like with like. YouTube shows a "typical retention" line for videos of similar length. Your goal is not 80%. Your goal is better than typical, consistently, because typical is what the algorithm already knows how to rank.

And when you are planning, try the outline test. Write the video as bullet points — one per beat. If you have four beats, you have a short video. If you have twelve, you have a long one. If you have four beats and you are planning twelve minutes, you are planning padding, and you know it. Cut before you film. It is cheaper.

The right length is the length where every minute earns the next one. Some ideas earn four minutes. Some earn forty. Your job is not to hit a number. It is to be honest about which kind of idea you have — and then to respect the viewer's time like it is your own, because in the attention economy, it basically is.