Seconds 0 through 3 of every clip are the visual or audio spike. This is the moment the algorithm decides whether to keep showing your clip to more viewers or to bury it. The decision is made on watch-completion rate at the 3-second mark. A clip that loses the viewer in the first 3 seconds is a clip the algorithm will not promote, regardless of how strong the rest of it is.

The visual spike options: a face on camera at maximum expression, a fast cut between two contrasting frames, a movement entering the frame, a color shift, a zoom-in or zoom-out. The audio spike options: a reaction sound, a music drop, a verbal hook ("you will not believe what just happened"), a sound effect that signals action.

Use both visual and audio spike if possible. The clips that retain the highest 3-second watch rate combine the two: a startled face accompanied by a "WHAT?" sound, a fast cut accompanied by a music spike. Stack the spikes. Earn the next 30 seconds.