The standard advice is volume. Post more, feed the algorithm, let the numbers sort it out. That is a lottery with extra steps, and it is why you can work this hard and still not know what worked.
So I stopped adding posts and read the ones I already had. Of the 51 with usable opening frames, I had never once cut within the first 3 seconds. 49 of 51 were close-ups. 50 of 51 used the same caption style. My skip rate still swung from 33% to 73%.
Fifty-one posts, and I had run the same experiment fifty-one times. That is the trap. More volume is more of the same experiment — and you cannot learn from a variable you have never changed.
I built a system that pulls every Reel I post, extracts what is actually happening in the opening frames, and tests it against how the post performed. The first thing it told me was not which hook wins. It was that I had never run an experiment in my life. Here is the receipt, and it is my own account.
Descriptive census of one account — mine. 53 Reels posted, 52 with metrics, 51 with usable opening frames. Skip rate is a figure Meta still labels “in development.” Not a prediction about yours.
Read the top row against the bottom row. The outcome swung by 40 points while the inputs sat still. There is no statistics in the world that can tell you whether cutting early helps, because I never once tried it. Neither, probably, have you — and I have not found a tool that mentions it.
That is the whole argument for posting less. Fifty more posts in the same shape teach you nothing new. Five posts that each change one thing you have never changed teach you something you can keep.
For weeks my tool told me: open with a hard stat and a named source. It looked airtight. My five best-saving Reels all did exactly that — four of them above a 5% save rate, the fifth at 4.8%.
Then I ran a permutation test across 50 posts. The rule failed at p = 0.82 and p = 0.96. It was noise. I had generated a confident rule by reading four winners.
That is how mine was built, and it is worth asking where yours came from. A model that reads your top posts will invent rules forever and never retract one. Only a test kills a rule. This thing is built to argue with itself, and the first thing it proved wrong was me.
Some of what gets sold in this category cannot be computed from what Instagram actually hands out. Here is the honest split for this one, and the cannot column is the one I would read first.
One more caveat I would rather you hear from me. On my own account, with 16 hypotheses tested and a strict correction applied, nothing currently survives. The two strongest effects are suggestive, not proven. I would rather tell you that than sell you a certainty I do not have.
Anyone whose next video should be informed by their last fifty. If you have an archive and no idea what is in it, that archive is the asset.
It works better the more you have posted. Under about 20 posts there is not much archive to read, and I would rather say so now than take your email under false pretenses.
Leave your handle and I will tell you when it is ready for you. Early access goes out in small batches, and the people who left a handle go first — because I can actually look at the account before I email.