Lock the ratio set before you cut hook variants

A hook test is only a test if one thing moves. If openings are 2.1 seconds in 9:16, 3.4 seconds in 4:5, and 1.8 seconds in 1:1, you did not test hooks. You tested a pile of files. The winner will be a ratio, a length, or a coincidence, and you will not know which.

The clips on this page are the same 9:16 Hook Studio examples — distinct openings, one portrait frame. That rail is already a ratio set. Production should look like that: one cut point, one ratio set, thirty openings.

How to merchandise a pack is using Versely hooks to promote a product. How to quote thirty openings as a supplier is selling hook volume. This post is only the constraint those two assume: lock the frame and the handoff before you cut variants, or the test cannot be read.

The two variables people smuggle in

Length. Every hook has to die at the same timestamp and hand the same body. If hook A is 2 seconds and hook B is 4, you also changed when the product appears, how much music plays, and where the caption sits. The platform will report a "better" hook that is actually a better pace. You cannot rerun that insight. Fix the cut point first: one number, on the wall, in the edit. Three seconds is a common one because hook rate is a first-seconds metric. The exact number matters less than the sameness.

Ratio. 9:16, 4:5, 1:1, 16:9 are different crops of the face, the product, and the burned-in line. A line that clears TikTok chrome in 9:16 sits under a Reels username in another crop, or loses the product at the edge. If you ship three ratios as three "hooks," the test is a layout test. Layout tests are valid. They are not hook tests. Run them later, on the winning line.

The pack agent renders 1080p portrait for a reason. Create a branded hook pack is a 9:16 object. If the media plan needs 4:5 and 1:1, those are a set you apply to every variant after the line is chosen — or, if the buyer requires all three in-test, you generate the whole thirty in each locked ratio as separate cells, still at the same cut point. What you do not do is mix.

What a locked production cell looks like

Write it down like a spec, not like a vibe:

• Body ID. One validated middle-and-end. Not "the new cut."
• Cut point. e.g. 2.8s. Every hook ends here. The body starts here.
• Ratio set. e.g. 9x16 only for round one. 9x16+4x5 only if the insertion order demands both, as parallel tests, named as such.
• Caption family. One. The hook line is already burned; do not also restyle it per variant.
• Count. Thirty openings, in four or five families, not thirty rewordings.

File names carry the cell. acme_offer_B02_question-H14_9x16 can be compared to …H15_9x16. …H14_4x5 cannot, except as a ratio check of H14. If the ratio is not in the name, the ads manager will not save you two weeks later.

This is the same discipline as not A/B testing two headlines on two different landing pages. Nobody would call that a headline test. Hook production forgets because the files are cheap. Cheap files still confound.

Where teams leak

"We'll recrop the best ones." Recrop after you have a winner. Recropping the whole batch to fill placements before anyone has a hook rate is how you spend the week in the export queue.

"Stories need 4:5 so those hooks can be longer." Then Stories is a different cell: different cut point, different ratio, different report. Do not dump it in the same campaign as the Reels 9:16 set.

"The model spat out 3.2s on this one." Trim to the cut point or reject the take. A beautiful extra second is how the body starts late and the test dies.

"We'll test 16:9 YouTube on the same lines." Horizontal is a different first frame. The face is smaller. The burn-in sits in a different safe zone. Same copy, new cell.

The hook pack is allowed to be portrait-only. That is not a limitation of the test. That is the test being possible.

Read only what you controlled

Hook rate — share of impressions still there a few seconds in — is the number the cell is built to move. CTR mixes thumbnail, targeting, and offer. Completion mixes the body you were supposed to hold still. If the body moved, you are not in this post anymore.

When a ratio must be compared, compare the same hook ID across ratios, not the winner of 9:16 against a random 1:1. One ID, two crops, one cut point. That is a layout test you can defend.

Until the cell is locked, do not generate the thirty. You will generate them again.

Safe zones travel with the ratio. 9:16 fights TikTok and Reels chrome at the bottom and the username at the top. 4:5 is a different crop of the same fight. 1:1 will clip a vertical burn-in you placed for 9:16. If the hook line is already burned, you cannot "move it in the ads manager." You recut. That is why the ratio is a production lock, not an export preset you pick when you are tired.

FAQ

Can I run 9:16 and 4:5 in the same ad set?

You can. You should not if you want a hook winner. Split by ratio, or accept that the platform will mix layout into the result. If the buyer demands both placements, two ad sets, same hook IDs, same cut point.

What if a placement rejects my locked ratio?

Then that placement is out of round one, or it is a new cell. Do not secretly recrop five of the thirty to sneak them in. The report will credit the wrong line.

Does the hook pack have to match the body's ratio?

Yes, at the cut. A 9:16 hook into a 16:9 body is a recut, not a variant. Generate or crop the body to the same set before you attach openings.

How many ratios belong in round one?

One, unless the insertion order names two placements as the job. Thirty openings already need a family structure to be readable. Thirty openings × three ratios is ninety files and a spreadsheet nobody will finish.