Built for: short-form teams, clippers and social leads who have produced a clip that worked and now have to produce fifty more, without knowing which part of it did the work.
One clip beat everything else you posted this month. You do not actually know why.
That is the normal position, not a lapse in attention. A clip that lands is nine decisions stacked on top of each other and delivered as a single object. The hook, the first frame, the feeling, the shape of the story, the person on camera, the edit, where the brand sits, what you asked for at the end, and who happened to be served it. All nine fired at once. One or two carried the result. The other seven came along for the ride.
Then comes the part that wastes the month. Teams do one of two things with a winner. They repost it until the platform stops serving it, or they brief "more like that" and every one of the nine variables moves at the same time. The next batch underperforms, nobody can say which change cost them, and the winner gets filed as a fluke.
This is the other approach. Pull the clip apart into its nine variables, work out which ones are load-bearing, hold those still, and move exactly one of the rest per round. Twenty variations out of one winner, each one telling you something you did not know before you posted it.
<aside> 💡
A winner is not a template. It is a hypothesis you have not tested yet.
The clip proved that some combination of nine things works on this audience. It did not tell you which combination. That is the whole job.
</aside>
Run this when you have one clip that clearly outperformed your baseline and you need to turn it into volume. Clearly outperformed means at least three times your account median on views, or a retention curve visibly flatter than everything around it. One good day is not a winner.
Do not run this on a clip that did well because of something you cannot repeat. A creator posting about a real event, a platform push, a duet from a large account, a news cycle. Those are borrowed conditions, not mechanisms. Copy the mechanism, not the weather.
Do not run it on a clip with fewer than a few thousand views on a network account, either. Below that the numbers are noise and every conclusion you draw from the test round will be wrong in a confident-sounding way.
Every short-form clip can be described completely by these nine. If you cannot fill all nine rows for your winner, you have not watched it closely enough yet.
| # | Variable | The question it answers | What to record |
|---|---|---|---|
| 1 | Hook | What makes someone stop inside the first second? | The exact words, spoken and on screen, and when each appears |
| 2 | Opening frame | What is on screen at 0.0s, before anything happens? | The still image itself. Framing, whether a human face is present, motion or stillness, colour weight |
| 3 | Emotional trigger | What does the viewer feel, in one word? | One word. Envy, outrage, tension, vindication, relief, disbelief, recognition, greed, schadenfreude |
| 4 | Story mechanism | What structural shape carries them to the end? | Near-miss, reversal, escalation, withheld reveal, before and after, POV, countdown, loop |
| 5 | Creator | Who is delivering it and what do they represent to this audience? | The archetype, not the name. Insider, outsider, expert, everyman, spectacle, faceless |
| 6 | Format | How is it cut, paced, captioned and scored? | Length, cuts per ten seconds, caption style, sound bed, aspect, whether it loops |
| 7 | Brand placement | Where does the brand appear, for how long, how loudly? | Timestamp of first appearance, total seconds on screen, treatment (native, overlay, verbal, none) |
| 8 | CTA | What are you asking for, and when? | Exact wording, position in the clip, and whether it is spoken, on screen, or caption only |
| 9 | Audience surface | Who was actually served this? | Account, platform, follower count, audience country split, day and time posted |
Variable 9 is the one teams forget, and it is the one that invalidates the most tests. A clip that worked on a 200K account with a 60 percent United States audience has not been tested on a 12K account with a Brazilian audience. Those are two different experiments.
Once you have the nine written down, every one of them gets exactly one of three states for the round you are about to run.
<aside> 🎯
The rule the whole system runs on.
Locked variables are held identical. Loose variables move freely but get recorded. Exactly one variable is Live per round. One.
</aside>