Look at five of your own retention curves side by side and something becomes obvious that no single curve shows: they share a shape. Almost all the loss happens in the first two seconds, the middle is comparatively flat, and whatever you changed in the edit further in barely moved anything.
That shape is the whole argument for treating the hook as a separate discipline. Fixing it moves the entire curve in a way no later edit can reach, and most effort spent on pacing in the middle is spent on the part that was already working.
The setup tax
Count the seconds between when your video started and when it began. Not when the first word was spoken: when the actual point arrived. Then compare that number with where the retention drop is. In most underperforming videos those two numbers are the same, which is a more useful diagnosis than anything about the content.
Introductions feel necessary to the person who made the video and are invisible to a viewer who has no context and no obligation to acquire any. The fix is almost always deleting the first sentence you wrote, which is uncomfortable precisely because it was the sentence that got you into the idea.
If the drop is at second two, no amount of faster cutting at second nine can reach it. Diagnose where the loss is before choosing a fix.
Why pure curiosity gaps stopped working
You will never guess what happened next used to work and now underperforms badly. Audiences have been trained out of it by several years of the payoff not being worth the wait, and a hook that withholds both the subject and the outcome is not intriguing, it is unparseable. A viewer who cannot tell what they are looking at leaves faster than one who is mildly bored.
What still works is naming the subject and withholding the outcome. I tested five cheap knife sharpeners tells you exactly what this is and gives you a reason to stay for which one won. That is a curiosity gap with a floor under it.
The hook that only exists in the audio
Most first views are silent. A spoken hook with nothing on screen has already failed for the majority of the people it reached, before any question of how good the line was. Hooks that work are said and shown, saying the same thing, which costs one line of on-screen text and is the cheapest fix on this list.
Testing a hook so the answer means something
Making four different videos and keeping the best one tells you which of four videos won and nothing about why, which means it cannot be applied to the fifth. A real hook test changes the opening line and holds everything else constant: same script from second four, same presenter, same length, same caption.
- Six variants minimum. Below that, normal variance in distribution swamps the difference you are measuring.
- One variable. Different hook and different length at once produces a winner you cannot learn from.
- Agree the stopping point before you look at the numbers. Deciding afterwards is how noise gets promoted to insight.
- Space them out. Six near-identical videos in three days compete with each other.
- Write down what you learned about hook style, not about this video. That sentence is the only durable output.
The part that used to make this impractical was producing six versions that differ in exactly one place, because humans cannot repeat themselves precisely enough to hold everything else constant. That constraint is the one that has genuinely changed, and it is why hook testing is now a Tuesday afternoon rather than a budget line.