Creators spend enormous effort optimizing thumbnails and titles — the lever that gets the click — while treating the edit as mere assembly. The data suggests this is backwards. The title wins the impression; the edit owns the session. And the session, not the click, is what algorithms reward.
Retention curves don’t care about your keywords. They respond to rhythm.
The Two-Second Rule and Its Exceptions
A working heuristic: no shot, idea, or visual pattern should persist past its welcome — and welcome expires in about two seconds of perceived stillness. But the rule isn’t “cut every two seconds.” It’s “change something every two seconds” — a cut, a zoom, an on-screen element, a music shift, a location change. The audience’s attention needs a reason to reset.
The Pacing Patterns That Retain
- Pattern interrupts at dip points. Every retention curve has predictable sag zones — the 15-second mark, mid-explanation, transitions. Insert visual change before the sag, not after the drop.
- Voice-over pacing > visual pacing. A dialogue rhythm that’s one beat too slow loses viewers faster than a static frame. Record tighter, edit looser.
- Sound design as the reset signal. A subtle riser or hit at a pattern break does more for attention than a new camera angle — viewers watch with their ears.
“The timeline is the only place the algorithm can’t help you. It’s also the only place that fully controls whether people stay.”
The Non-Obvious Edit
The highest-leverage edit is subtraction: cutting the ‘throat-clearing’ sentence that opens every paragraph, tightening the pause before the punchline, removing the transitional sentence the script didn’t need. Retention lives in the words you removed.
Editing patterns, cut-density benchmarks, and timing formulas in The Editing Craft Compendium.
The edit decides what the feed gets to show.