Open any short-form retention graph and you will see the same cliff: a vertical drop somewhere between second one and second three, then a gentle slope for whoever survived. Everything you spend on production is decided by that cliff.
The cliff is a question, not a hook
Viewers are not asking “is this good?” They are asking “is this for me?” A hook works when it answers that question faster than the thumb can move. Beauty does not answer it. Specificity does.
Four openings that keep working
- The named audience. “If you run a clinic with two chairs and no front desk…” — narrow enough that the right person cannot scroll past.
- The contradiction. Show the result before the setup. It creates a gap the viewer wants closed.
- The mid-sentence start. Begin as though the conversation was already happening. It reads as native, not as an ad.
- The visual anomaly. One frame that does not belong in a feed. Use sparingly; it fatigues fast.
What we stopped doing
Logo stings at the top. Slow establishing shots. Music-first openings with no voice. Captions that repeat the voiceover word for word instead of adding information.
Cut for the platform, not the edit suite
The same interview can produce a 22-second vertical cut, a 90-second feed piece and a three-minute YouTube version. They are not crops of each other; they are different edits with different first seconds. Budget for that.
Test the hook before you finish the film. We cut six openings against the same body, spend a small amount on each, and only then finish the winner properly.