The 8 minute rule came from an ad policy, not from ranking
The legend has a birth certificate. From 2015, mid roll ads were only available on videos of 10 minutes or longer, so a whole generation of creators learned to write to ten minutes regardless of what the topic deserved. In July 2020 YouTube lowered that threshold to 8 minutes, and the folklore simply slid down two minutes with it.
The cutoff is real and brutal in exactly one direction: a video of 7:59 gets zero mid roll slots, a video of 8:00 gets them, and there is no partial credit. That is a billing rule about advertising inventory. YouTube has never published video length as a ranking factor, and no public statement has tied recommendation to crossing eight minutes.
What actually happens is the opposite of what people expect. The padding you add to cross the line lowers retention, and retention is the signal the system genuinely reads. A 6 minute video holding 65 percent delivers 3.9 minutes per view. Stretch it to 8:10 with a long intro and a recap, watch retention fall to 40 percent, and you now deliver 3.3 minutes. You traded a real signal for one ad slot.
What YouTube optimizes is session time and satisfaction, not raw minutes
The system is not asking how long your video is. It is asking whether serving that video to that viewer produced a good session: did they keep watching, did they watch something else afterwards, did they come back tomorrow, did they dismiss the recommendation. Duration is one input into that question, never the answer to it.
Here is where two metrics get confused. Percentage viewed compares you against videos of similar length. Minutes watched is the absolute fuel behind watch hours and ad inventory. A 25 minute video at 45 percent delivers 11.25 minutes per view, while a 6 minute video at 70 percent delivers 4.2. The long one produces 2.7 times the watch time from the same single view, and carries two or three ad breaks while doing it.
Percentage still matters, because YouTube compares you inside your length band and not against the whole platform, where average retention sits between 35 and 45 percent. Read the 2026 bands below and notice that 45 percent on a 25 minute video is above its band, while 70 percent on a 6 minute video merely sits inside its own.
- Under 5 minutes: 65 to 75 percent is strong, and below 50 percent the hook or the topic missed
- 5 to 10 minutes: 50 to 60 percent is strong, the band where most tutorials and reviews live
- 10 to 15 minutes: 40 to 50 percent is strong, and this is where padding shows up as a cliff
- 15 to 30 minutes: 35 to 45 percent is strong, and absolute minutes per view start to dominate
- 30 to 60 minutes: 25 to 35 percent is normal, and completion rate stops being a useful goal
- Over 60 minutes: 20 to 30 percent is normal, so judge the video on minutes per view alone
The length that works, niche by niche
There is no universal number, there is a number per type of promise. The ranges below hold up in 2026, and each one comes with the reason attached, because the reason is what lets you break the range intelligently when a specific topic deserves it.
Read them as the shape of the viewer's intent, not as a quota to hit. Someone searching how to fix a leaking tap wants to be gone in five minutes with dry hands. Someone clicking on the collapse of a two hundred year old bank has cleared their evening and will be annoyed if you wrap up in eight minutes.
One channel can legitimately live in two bands at once. A finance channel publishing a 4 minute reaction to a rate decision and a 22 minute breakdown of a balance sheet is not being inconsistent, it is serving two intents. What kills channels is picking one number and forcing every topic through it.
Living in two bands is easy to say and expensive to do, because by hand it means two different production routines running in the same week. In FalconVid the target duration is a setting on the project rather than a workflow, so the 4 minute reaction and the 22 minute breakdown come out of the same pipeline on the same calendar. And when the two intents genuinely deserve separate homes, each channel gets its own calendar, its own visual identity and its own language, 5 channels on Pro and 50 on Scale, so serving two audiences stops meaning splitting your week in half.
- Tutorial and how to: 4 to 8 minutes for a single task, 12 to 20 for a full workflow, because the viewer has a problem and a clock running
- Product review: 8 to 14 minutes, long enough for setup, real tests and a verdict, short enough that a buyer deciding today finishes it
- News and commentary: 3 to 6 minutes, since the value decays by the hour and depth was not the reason for the click
- Lists and curiosities: 8 to 15 minutes, because each item resets the viewer, so 45 to 55 percent retention is normal here
- Gaming, edited: 12 to 25 minutes, while unedited playthroughs run 40 minutes and up and are judged on minutes, not percentage
- Documentary and story: 18 to 40 minutes, carried by narrative tension, where 35 to 45 percent at the 30 minute mark is excellent
- Finance and business: 10 to 18 minutes, and at $20 to $40 RPM in the United States two breaks here outearn four in a low CPM niche
- Podcast and interview: 45 to 120 minutes, an audience that listens rather than watches, so 20 to 30 percent is normal
The test that settles it: the video lasts as long as the title promised
Every title signs a contract, and every contract has a duration attached. How to change a bike tyre promises about five minutes. Why Kodak invented the digital camera in 1975 and still died promises twenty five. The viewer arrives with an internal stopwatch already running, and when you run past it they leave feeling cheated even when the content was good.
Applying it is mechanical. Before writing anything, put the promise of the title into one sentence, then ask how long it takes to deliver exactly that sentence and nothing else. That is your target length. The script exists to fill that window, not to reach a threshold that only matters to the ad server.
The cut is where it becomes real. After the edit, walk the timeline in 60 second blocks and ask what each block adds to the promise. Anything that only adds context or survives because you liked the joke goes. Doing this honestly for the first time usually removes 20 to 35 percent of the running time and lifts percentage viewed by 8 to 15 points.
The reason almost nobody runs that 60 second audit is when it happens: at the end of a long day, with the file already exported and the slot tonight. In FalconVid the promise sets the length before anything is produced, because you choose the target duration on the project and the script is written to fill exactly that window, so the padding never gets generated in the first place. And when a first cut still runs long, the Studio is where you shorten the intro, drop a scene or trim the ending, which is the same audit done in a few minutes rather than at midnight.
- Write the promise of the title as one sentence before the script, and let it set the target length
- Kill the recap: whoever clicked already knows, and 40 seconds of preamble costs you the 30 second checkpoint
- Delete the origin story, because nobody searching for a fix wants the history of the problem first
- Cut any block that would still make sense if it were deleted, since anything removable was padding
- If the honest answer is 6 minutes, publish 6, because one ad slot is not worth 20 points of retention
- If the topic genuinely needs 30 minutes, do not compress it into 10, because rushing reads as low value
Where retention actually dies: three very specific places
The first 30 seconds is the biggest hole in almost every video. Typically 20 to 40 percent of viewers are gone before the 30 second mark, and in a weak video it is over half. This is the highest leverage stretch in the entire timeline, far above length. If your 30 second checkpoint sits under 60 percent, no change in duration will rescue that video.
The second cliff lands between minute 2 and minute 3, when the energy of the intro has burned off and actual substance has to start. A healthy video loses 8 to 15 points across that stretch, a padded one loses 25 or more. This is precisely where material added to reach eight minutes sits, which is why padding shows up as a visible ledge rather than a gentle slope.
The third is the dragged ending, the last 10 to 15 percent of the timeline: the outro, the thanks, the if you enjoyed. Retention there is not free. It drags the average down and pushes the viewer to leave the session instead of rolling into your next video, which is exactly what the recommendation system is watching. End when the promise is fulfilled and let the end screen work in under 15 seconds.
Two of those three holes are structural, which means they can be closed by default instead of by willpower. FalconVid writes every script hook first, so the promise leads and there is no branded intro burning the 30 second checkpoint, and the dragged ending never gets generated because the video stops when the promise is fulfilled. The middle cliff is the one that stays yours to watch. If minute 2 to 3 keeps collapsing across a whole cluster, the topic was thinner than the title claimed, and no pipeline can rescue a promise that was never worth eight minutes.
Mid roll: where to place breaks and what they honestly change
Automatic placement is the default and in 2026 it is reasonably good, inserting breaks at natural pauses in speech. Manual placement still wins when your video has real chapters, because a break landing in the middle of a sentence is what makes people close the tab. The ad is tolerated, the interrupted thought is not.
The practical rules are simple. Do not put the first break before minute 3 or 4, keep roughly one break every 4 to 6 minutes, and place it right after a payoff and never immediately before one. A 12 minute video carries 2 breaks comfortably, a 20 minute video carries 3 or 4, and anything denser than one every 3 minutes costs more retention than it earns.
The money, in honest ranges. Long form RPM in the United States runs $6 to $14 with a single break and $10 to $18 on a well retained video carrying two or three. Finance and B2B run $20 to $40, Brazil and Latin America $0.50 to $1.50. On 100,000 views the gap between one break and three is roughly $400 to $600 in the United States and $30 to $60 in Latin America.
Sleep, ambient and compilation: a completely different sport
Everything above collapses the moment you upload a 10 hour rain sound video. Percentage viewed is meaningless there: 3 to 8 percent is normal and perfectly healthy, because the viewer fell asleep, which is the exact outcome they came for. The satisfaction proxy in this category is whether they return tomorrow night, not whether they reached the end.
The absolute arithmetic is startling once you run it. Four percent of a 10 hour video is 24 minutes per view, roughly six times what a 6 minute tutorial at 70 percent delivers. Ten thousand views at that rate is 4,000 hours of watch time, the entire Partner Program requirement for anyone accepted until 31 January 2027, generated by a single upload. From 1 February 2027 a new channel needs 8,000 qualified public watch hours instead, which the same video clears with twenty thousand views.
The catch is on the revenue side. Music, ambient and sleep are low CPM categories, often $1 to $3 in the United States and under $1 elsewhere, because advertisers are not eager to buy an unconscious audience. Mid rolls there wake people up and generate dismissals, so breaks should be far apart or absent. Compilations sit in between at 20 to 45 minutes, played in the background and judged on minutes, and every segment has to stand alone because viewers jump around.
Once you pick a target length, it is your choice on the calendar
The reason most channels never test length properly is production cost. A real test means publishing ten videos at 8 minutes and ten at 18 in the same month, same niche, same thumbnail style, then reading the retention graphs. At $50 to $300 per freelance edit, or six hours of your own evening per video, nobody runs that test. They just repeat the number they read in 2015.
This is the part FalconVid exists for. You approve a calendar (which days, how many videos per day, the exact time of each in your timezone) and you choose the target duration. The pipeline then researches, writes, narrates in any of 63 languages, selects images and clips per scene, edits, captions and publishes on its own across YouTube, Instagram, TikTok, Rumble and Facebook. The duration is your choice, set per project: if your niche calls for 6 minutes you publish 6, if it calls for 25 you publish 25.
The economics are what make the experiment possible, as long as the numbers come with the quality mode attached, because the mode moves the bill far more than the plan does. A 12 minute long video costs 1,008 credits in economy, 8,676 in balanced and 26,760 in premium. Starter is $47 a month with 15,000 credits, which is 14 videos if you stay in economy the whole month or exactly one if you shoot everything in balanced, so the realistic working plan is around 10 to 12 videos a month mixing economy with one in balanced. Pro is $97 with 30,000 credits, 5 channels and 5 concurrent generations, up to Scale at $997 with 50 channels and 50 concurrent generations, and every plan carries every creation feature with a 7 day guarantee.
So the honest ending has two answers rather than one. Producing by hand, the length question simply stays open forever, because a real test costs twenty finished videos inside one month and nobody has twenty videos of spare weekend, which is precisely how a 2015 number survived a decade past its expiry date. Producing in parallel, with videos generated side by side and the target duration set per project, that same test is just an ordinary month of publishing: ten at 8 minutes, ten at 18, same niche, same calendar, then read the graphs. The right length for your niche is knowable. It was only ever the production cost that kept it a matter of opinion.