Write the title and thumbnail before the script

On YouTube, packaging is the strategy and the script is the delivery. If you write the video first and then try to package it, you end up with a title that describes the video accurately and interests nobody. Decide the promise first — the specific curiosity gap a viewer clicks on — then write a script that pays it off.

The first thirty seconds should restate that promise in different words, show a glimpse of the payoff, and explain the shape of the video. Do not deliver the answer yet, and do not spend that window on your channel, your sponsor, or your week. Absolute view duration is measured from second one, and the audience retention graph will show you exactly where the intro ran long.

Structure for the retention graph

Long-form retention decays smoothly and drops at seams. The seams are transitions, so script them: a one-line signpost before each new section keeps viewers who would otherwise leave at a hard cut. Roughly every ninety seconds, plan a pattern interrupt — a B-roll sequence, a screen recording, a location change, a graphic. Interrupts do not have to be dramatic, they just have to be different.

Write in chapters, and name them the way someone would search for them. Chapters give you timestamps for the description, and timestamps give the video a second life as a reference. If you run past eight minutes you unlock mid-roll ads, but only place them where the script has a natural beat break — a mid-roll dropped mid-sentence is a measurable retention cliff.

The last twenty seconds belong to the end screen. Do not summarise; a summary tells viewers the video is over. Instead hand them the next video that follows logically from this one, and say why it follows.

Common YouTube script mistakes

The cold open that is really a warm-up. "Hey guys, welcome back to the channel, before we get started" costs you the exact seconds the browse feed uses to decide whether to keep recommending the video.

Reading the script. Long-form viewers notice performance more than short-form viewers do, because they are watching for longer. Script the beats and the key lines, then talk through the rest.

Padding to hit ten minutes. Watch time comes from retention, not runtime; a tight six minutes beats a slack twelve every time.

Ignoring the description. Two lines of real context plus chapter timestamps is a small amount of work that keeps paying off in search.