Long-Form Narrative Structure

Getting somebody to minute two is a hook problem. Getting them to minute twelve is a structure problem, and it's a different job entirely.

Start by choosing a structure that holds attention by design. Discovery to revelation, where the video is the process of finding something out. Then versus now, where the gap between two states is the story. Mistake to mastery, where somebody gets it wrong long enough that getting it right means something. Pick one before you write. A shape with tension built in does half the retention work, and a shape without it can't be rescued by good sentences.

The shape to avoid is the one school taught you. Thesis, three supporting points, summary. That's an essay, and read aloud it's flat. Nothing builds, nothing's at stake, nothing to wait for. You want an arc.

Then assume attention decays, because it does. Nobody sits still for fifteen minutes on willpower. So you interrupt the decay on a schedule: a major pattern interrupt every sixty to ninety seconds. A hard cut to different footage, a shift in the music, a change of pace, a visible change in what kind of video this suddenly is. Which one matters less than it landing before the drift, not after.

Give them a reason to still be there in three minutes, too. That's what open loops do. Promise something specific and valuable, then deliberately go elsewhere before you deliver it. Say plainly you'll show them the thing that ended the case, then spend two minutes on how it started. An unfinished thought sits in the mind far more insistently than a finished one, and people stay to close it whether they've decided to or not.

Order your material so it rises. The instinct is to lead with your best point, and inside the body of the video that's wrong. Stack points in increasing order of value, so the viewer's brain keeps expecting this to get better. Spend your best material at minute three and everything after is a decline they can feel.

Now the beat-to-beat mechanic, and this one rule will do more for your scripts than anything else here. Between any two story beats, the connector is "therefore" or "but". Never "and then." That's the South Park rule. A chain of and-thens isn't a story, it's a list of events in date order, and lists have no pull. This happened, therefore that happened. She won the case, but the money had already gone. Every beat caused by the last one or in conflict with it. Go through a finished script, find every and-then, force it into one of the other two.

There's a set of cheap moves that lift average view duration, and they're worth writing in deliberately. Transition sentences buried mid-paragraph, so the audio never rests at a tidy place to leave. Re-hook phrases dropped exactly where interest sags. But that isn't the worst of it. It gets stranger. A face-cam clip cut in. A change of voice, or a swap in the music. And strip the audio gaps out, because the silence between sentences is where a viewer decides they're finished.

Feed the eyes as hard as the ears. Write in a concrete visual reference every twenty to thirty seconds of script, so there's always a real image to cut to instead of another stock city skyline. And pack the writing with specifics. Actual dates, names, places, numbers. Specificity is what makes narration sound researched rather than generated.

Long-form runs slower than a short, and people read that as permission to coast. The pacing is slower. The value beats aren't. Something has to keep arriving or the viewer drifts.

And if the topic allows, don't settle the central mystery. Lay out the folklore, offer practical explanations, resolve nothing. They keep watching for an answer, then argue in the comments.

Cause and conflict. That's the whole engine.