Hook rate: the metric that isolates the first seconds

Every video asks two questions in sequence. The first is "is this for me", answered in under three seconds. The second is "is it worth continuing", answered across the rest. Hook rate measures the first one, alone.

That separation is what makes the metric valuable. Average retention and watch time blend both questions into one number, and therefore do not tell you whether to rewrite the opening or re-edit the middle.

In practice the calculation is: how many people were still watching at the third second, divided by how many started. Some panels give it directly; where they do not, the retention curve provides it, because it is the curve's height at that point.

Re-editing a whole video costs hours. Changing the first three seconds costs minutes, and it is the change with the largest measured effect per unit of effort.

The test that works is boring in its simplicity: take a video with a hook rate below your median, change only the opening and republish a few days later. With everything else held constant, the difference is attributable.

The mistake that voids the test is changing two things. Altering hook and cover, or hook and caption, produces a result that teaches nothing, because you cannot know which change caused the effect.

• Intros or animated logos. A second of branding is a second of audience.
• Greetings and self-introductions. Someone who has not decided to stay does not care who you are.
• A static opening. A person standing still facing camera loses to anything happening.
• A generic promise. It works for everyone, therefore it speaks to nobody.
• Illegible on-screen text. A large share of consumption is sound off; if the promise only exists in speech, it does not exist.
• Quiet audio. A video quieter than the rest of the feed is skipped before the first word.

There is a trap that shows up often: an excellent hook with a poor hold rate. The curve makes the pattern obvious: almost nobody leaves in the first seconds and everybody leaves right after.

That means the opening promised something the content did not deliver. On top of wasting the audience you earned, it generates negative signal, because the platform observes that people left early after having stayed.

The correct pair to track is always hook rate and hold rate together. One measures whether it grabbed, the other whether it held.

• I track hook rate separately from average retention.
• I test by changing only the opening, never two variables.
• I check that the promise works with the sound off.
• I cut intros, greetings and self-introductions.
• I always read it alongside hold rate, so I do not confuse hook with content.