YouTube retention: what the first 60 seconds of a long video must do
The opening minute decides whether the other fourteen get watched. Four jobs it has to do, the editing rules behind them, and a worked example shot by shot.

YouTube retention is decided early. Open the audience retention graph for any long video and the shape is the same: a steep drop at the start, then a slope that flattens. Everything after the first minute is watched only by the people who survived it. So the opening gets more editing attention per second than any other part of the video, and this piece explains what to spend that attention on.
Read your own graph first
In YouTube Studio, the retention graph for a video shows the percentage of viewers still watching at each moment. Three features matter.
- The opening drop: how far the line falls in the first 30 to 60 seconds, which tells you whether the video delivered what the title and thumbnail promised
- Dips: sudden falls later on, which mark a moment where people got bored or confused and left
- Bumps: small rises where viewers rewound, which mark something worth doing again
Compare your videos with each other, and with YouTube's own "typical" line in Studio, not with numbers quoted by other creators. Retention varies with length, niche and traffic source, so someone else's percentage tells you little about your own.
Four jobs for the first minute
Confirm the click
The viewer clicked on a title and a thumbnail. The first five seconds must show them they are in the right place: the same subject, the same stakes, ideally the same image. A logo sting, a greeting or a channel intro delays this, and each second of delay costs viewers.
State the stakes
Say what happened and why it matters, in concrete terms, before any background. Numbers, names, dates and places do this better than adjectives. "At least 6,000 people died" does more work than "a terrible tragedy".
Open the central question
Give the viewer one question that the whole video will answer, and make it a question they cannot answer themselves from the title. This is the loop that holds the video together.
Promise the route
One sentence on what the video will cover, so the viewer knows the answer is coming and roughly how. Then start. Do not list chapters aloud.
What does not belong in the first minute: your channel name, a request to subscribe, an apology for audio quality, or three sentences of throat-clearing about how interesting the topic is.
Visual density: the editing rules
A strong script can still lose people if the screen stays still. These are the rules we edit to, and the reasoning behind each one.
| Rule | Why it works |
|---|---|
| Strongest image and text from frame one | The viewer decides within seconds, so the best material cannot be saved for later |
| 10 to 12 visuals in the first minute | A change every five or six seconds gives the eye a reason to stay while the story is still being set up |
| A new visual or text card every 5 to 8 seconds through minute three | The early slope of the graph is steepest, so the pace stays high until viewers are invested |
| A new open question every 60 to 90 seconds | Each answered question releases some viewers unless another is already waiting |
| Subscribe ask at about 40 percent in | The people still watching are the ones likely to say yes, and a peak in the story carries the ask |
| No long stretches of plain footage | Unbroken footage under narration reads as filler, and dips cluster there |
A "visual" here means a real change: a new shot, a map, a document close-up, a name card, a number on screen. A slow zoom on the same photograph for 20 seconds counts as one visual, however much it moves. Text on screen should carry information the narration does not repeat word for word, such as a date, a place, a figure or a quoted line.
Density is not speed. Narration stays at a normal documentary pace. The picture changes underneath it.
A worked example: a history script
Take a 14-minute video on the Galveston hurricane of 1900. A weak opening would start with the founding of the city and reach the storm in minute three. Here is the first minute built to the rules above, with about 150 words of narration and 11 visuals.
| Time | Narration | On screen |
|---|---|---|
| 0:00 | On September 8, 1900, the richest city in Texas was under water by nightfall | Archive photo of wreckage with a slow push in, date line: Galveston, September 8, 1900 |
| 0:06 | At least six thousand people were dead | Text card over a debris field: At least 6,000 dead |
| 0:10 | It is still the deadliest natural disaster in American history | Wide archive photo of flattened city blocks |
| 0:15 | And the city had been told it could not happen | Close-up of an 1891 newspaper column, one line highlighted |
| 0:20 | Nine years earlier, the local weather chief had written that the idea of a hurricane seriously damaging Galveston was an absurd delusion | Portrait with name card: Isaac Cline, US Weather Bureau, then the quoted words on screen |
| 0:30 | The highest point on the island stood less than nine feet above the sea | Map of Galveston Island, elevation marker: 8.7 feet |
| 0:36 | The water that came ashore that night was more than fifteen feet deep | Animated comparison: island height against storm surge |
| 0:42 | There was no seawall | Photo of the open beachfront before the storm |
| 0:45 | Warnings from forecasters in Cuba had been dismissed | Map of the Gulf with the storm track drawn from Cuba toward Texas |
| 0:50 | So how did a city of nearly 38,000 people meet the storm with almost no notice, and who decided they did not need any | Street scene from 1900, text card: Who decided |
| 0:58 | This is the story of the hours before the water arrived | Title card over the first shot of the next section |
Check it against the four jobs. The click is confirmed in the first six seconds with the place, the date and the outcome. The stakes are stated twice, in the death toll and the ranking. The central question arrives at 0:50, and the viewer cannot answer it from the title. The route is promised in the final line. The city's founding, its cotton trade and Cline's biography all still appear in the video, later, once the viewer has a reason to care about them.
Every image in that table exists in public archives, which matters, because a hook you cannot source is a hook you cannot use. See public domain footage for YouTube for where to look.
Open loops after the hook
The central question should not be answered until near the end. Smaller questions keep the middle alive. Roughly every 60 to 90 seconds, the script should raise something new before it finishes resolving the last thing.
In the Galveston script, the loops might run like this: why the Weather Bureau distrusted the Cuban forecasts, what Cline saw at the beach that morning, why the bridges to the mainland were already gone by afternoon, what happened to the people who sheltered in the largest buildings. Each one is set up a paragraph before the previous one closes.
A loop must be paid off. Teasing an answer and never giving it works once and then appears in the comments for the rest of the channel's life.
Where the subscribe ask goes
Asking in the first minute interrupts a viewer who has no reason yet to say yes. Asking in the last minute reaches only the few who are still there. The better place is about 40 percent of the way in, at the most intense moment of the first half. In a 14-minute video that is between minute five and six.
Keep it to one sentence, tie it to the content, and move on. In the example, it falls as the water reaches the city: one line saying that if this is the kind of history the viewer wants more of, subscribing is how they will see the next one. Then back to the story mid-scene. Back it up with a pinned comment so the ask exists in text as well.
Check the opening before you publish
- Play the first 15 seconds with the sound off and ask whether the subject is clear from the pictures and text alone
- Count the visual changes in the first minute, and add more if there are fewer than ten
- Find the sentence that states the central question, and move it earlier if it sits after the one-minute mark
- Delete any sentence in the opening that could be removed without the viewer noticing
- Confirm the first frame matches the promise of the thumbnail
The script decides most of this before any editing starts, which is why length and structure need planning together. How long should a YouTube video be covers the length side, and the title generator helps make sure the promise the hook has to keep is a good one.
These pacing rules are also what JetMotion's templates apply automatically when they build a shot plan from a script. The how it works page shows the plan you see before a render. Edited by hand or by software, the test is the same: watch your own first minute as a stranger would, and be honest about the second you would have left.


