Hooks that hold: scripting the first three seconds

A viewer decides whether to stay inside 1.5 seconds. The platform decides whether to keep pushing you inside the first three. Everything you build after that, the pacing, the payoff, the CTA, is downstream of a hook that already worked. We shoot every piece with three hook variants and kill the two with the lowest 3-second retention after 24 hours. The surviving hook becomes a template.

Why the first three seconds

The platforms are not shy about this. Internal retention curves weight the opening heavily:

TikTok scores 3-second completion against a creator baseline. A hook that holds 85%+ gets pushed into unfamiliar audiences.
Reels uses "watch time per impression" as the dominant push signal. The denominator inflates fast; the opening protects it.
Shorts benchmarks against similar creators in the same vertical. A below-median first second caps distribution for the next 72 hours.

Viewer habit reinforces the algorithmic weight. The thumb is already moving when the frame loads. The hook has to interrupt the swipe, not request it.

Five hook structures that consistently outperform

Across our last 600 posts, these five hook shapes land in the top quartile of 3-second retention more than three times out of four.

The pattern break. Open on something visually unexpected: a cut mid-motion, a close-up on an object with no context, an on-screen word that contradicts the thumbnail. The viewer stays to resolve the mismatch.
The contrarian claim. "Posting daily is killing your reach." The viewer stays to argue with you.
The specific number. "I ran this ad for $87 and it made $4,210." Round numbers read as hypothetical; odd numbers read as receipts.
The cliffhanger question. "Why do most $10k launches stall in week two?" The viewer stays for the answer.
The before and after jump cut. Second one shows the end state, second two shows the start state, second three starts explaining the gap. Classic because it works.

The failure mode shared by all five is softening the opening with throat-clearing: "So, I was just about to post this, and I thought..." Every filler word before the hook costs you a percentage of retention.

Writing the opener

Three rules that save most hooks:

Lead with the sharpest word of the sentence. Not "You might think," but "Think." Not "There is a thing I noticed," but "Noticed."
Cut filler. "So," "basically," "I was gonna say," and "to be honest" are retention taxes.
Match the thumbnail to the first frame. If the thumbnail promises a specific result, the first frame must deliver the setup for that result. A mismatch is why viewers bounce at 1.5 seconds even when the hook is strong.

Testing hooks like a system

Shoot three hook variants per video. Same body, different openings.
Publish all three (one per day, same window if possible) or run them as paid tests on one platform.
Keep the winner, template it, kill the losers. After 30 days you have a library of 10 to 15 working hook shapes specific to your audience.
Re-test quarterly. Audience taste drifts; a hook that worked in Q1 can underperform by Q3.

What good looks like

For an established creator in a scaled niche, our target is a 3-second retention above 78% on every published piece. Below 65%, we treat the hook as broken and reshoot. Hooks are not creative garnish. They are the throttle on everything else you do.