The two questions that decide length
"How long should my video be" is unanswerable as asked — but it decomposes into two questions that are:
Question 1: How much video does the promise need? Every video makes a promise (the title, the thumbnail, the hook). The correct length is the shortest version that fully pays that promise off. A "5 mistakes" listicle promises five delivered mistakes — maybe 9 minutes. "The complete beginner's guide" promises completeness — maybe 25. The promise sets the floor; padding past it is where retention goes to die, and cutting below it is where satisfaction does.
Question 2: What is this video for? Search traffic, browse/suggested reach, watch-hour accumulation toward monetization, subscriber conversion, or revenue per video — each goal tilts the length decision differently, and we'll take them in turn below.
Everything else — niche norms, the 8-minute threshold, competitor lengths — is input to those two questions, not a replacement for them.
The retention math (why "longer = more watch time" is half true)
YouTube's recommendation systems optimize heavily for watch time, which tempts creators toward longer videos. The math has a catch:
Watch time per view = length × average retention. Stretch a 10-minute video to 20 and retention doesn't stay at 50% — it collapses, because the added minutes are your weakest material by definition. A 10-minute video at 55% retention delivers 5.5 minutes per view; the bloated 20-minute version at 22% delivers 4.4 — less absolute watch time, plus worse retention signals, plus a weaker impression on every viewer who left annoyed.
But genuinely earned length compounds. The same 20 minutes at 45% — because the material actually sustains it — delivers 9 minutes per view, nearly double the tight 10-minute video. This is why documentary channels and video essays punch so hard on watch hours: length that retention survives is the single most powerful watch-time lever on the platform.
So the honest rule: length amplifies whatever your retention already is. Earn length with material; never manufacture it with padding. Your retention graphs tell you which one you've been doing — more on reading them below.
The 8-minute threshold (and what it's actually worth)
One structural fact does create a genuine breakpoint: videos 8 minutes and longer are eligible for mid-roll ads, which typically raises revenue per view meaningfully in monetized channels. Three honest notes on it:
- It's a threshold, not a target. An honest 7-minute video stretched to 8:10 for the mid-roll usually loses more in retention (and therefore future impressions) than the extra ad earns. The threshold rewards videos whose material naturally clears 8 minutes.
- It compounds with niche RPM. In high-RPM niches (finance, business, software), clearing 8 minutes matters considerably more per view than in entertainment niches — factor it into format decisions, not per-video padding. The earnings calculator shows the niche spread.
- Pre-monetization, it's irrelevant to revenue and relevant to watch hours — an 8–12 minute format simply accumulates the 4,000 hours faster than a 4-minute one at equal retention, which is a legitimate reason growing channels drift longer.
Length by format: the working ranges
Ranges, not rules — the promise still governs. But these are where each format's promises typically land:
| Format | Working range | Why |
|---|---|---|
| Tutorial / how-to | 5–12 min | Searchers want the solution; completeness beats brevity only until the solution is delivered |
| Listicle | 8–15 min | Item count × ~90–150 sec per item; flagged-item structure carries the back half |
| Video essay / commentary | 12–30+ min | The format's promise is depth; short essays underdeliver their own genre |
| Documentary-style | 20–45+ min | Highest watch-time ceiling on the platform when research earns it |
| Product review | 8–15 min | Verdict + evidence + honest caveats; past 15 usually means two videos |
| News / reaction | 5–10 min | Timeliness content decays; velocity beats polish |
| Vlog / personality | 8–15 min | Carried by presence, capped by narrative shape |
| Podcast / conversation | 30–90+ min | Different consumption mode (background listening); chapters do the length management |
Two refinements worth more than the table. Search-intent videos should run shorter than browse-intent videos on the same topic — a searcher wants resolution, a browser wants an experience. And your niche's norms are a prior, not a law: check the top 10 results for your target topic, note their lengths, then ask whether they're that length because it works or because everyone copied everyone. Gaps in the norm (the 8-minute version of a topic everyone covers in 25, or vice versa) are positioning opportunities.
Shorts: the length question nobody frames right
Shorts now allow up to 3 minutes, which changed the question from "how do I fit 60 seconds" to "how short should this Short be." The feed's physics still favor brevity: completion rate and rewatches are the Short's retention currency, and both fall as length rises. The working pattern — under 60 seconds for velocity plays (subscriber growth, reach), longer Shorts only when the payoff genuinely needs the runway. Structure matters more than duration either way: promise in the first 1–2 seconds, delivery in fast beats, an ending that loops back to the opening. And remember the asymmetry from the monetization math: Shorts build subscribers, not the 4,000 watch hours — length strategy for Shorts is audience strategy, not hours strategy.
Finding your number: read your own graphs
Every generalization above loses to your channel's own data. The method:
- Pull your last 10–15 videos' retention graphs and note where average retention crosses 30–35% — the point past which you're talking to a small minority. If that consistently happens at minute 7 in your 14-minute videos, your audience is telling you your format's real length.
- Compare watch time per view across your length experiments, not retention percentage — percentage punishes longer videos cosmetically even when they deliver more absolute minutes. Watch time per view is the honest metric.
- Diagnose the drop shape. A cliff in the first 30 seconds is a hook problem, not a length problem. A steady slope is normal decay. A mid-video ledge marks a section viewers skip — cut that section, and the "length problem" often disappears without the video getting shorter.
- Script to the target, then cut. Once you know your number, write to it deliberately — at ~145 spoken words per minute, a 9-minute target is a ~1,300-word script (the full pacing table is here). Scripts that hit length by design don't need padding or panic cuts in the edit.
That last step is where length stops being a mystery: channels with erratic lengths almost always have erratic scripts — writing until they run out of things to say, then discovering the length in the edit. Scripting to a deliberate length target, with retention structure carrying every minute of it, is how the decision becomes repeatable. (It's also exactly how UpTube writes: you set the target length, and the script agent builds retention structure to fill precisely that runtime in your voice — from 60-second Shorts to 45-minute deep dives. The free plan will draft your next one at whatever length your graphs just told you.)