The hook, body and CTA are three separate experiments, not three sections of one script. The hook is measured by hook rate, the body by hold rate and CTR, and the CTA by conversion rate. Change one part at a time, hold the other two steady, and the result tells you which part is throttling performance.
The three parts, measured separately
A hook-body-CTA structure is a testing model first and a script format second. When you treat each part as its own experiment, you stop asking "did the ad work" and start asking "which section lost them". Here is how we split it.
| Part | Typical length | Job | Diagnostic metric |
|---|---|---|---|
| Hook | 0 to 3 seconds, occasionally 0 to 5 | Stop the scroll and buy the next 3 seconds | Hook rate: 3-second video views divided by impressions |
| Body | 3 seconds to the final 5 to 8 seconds | Deliver the offer, show proof, handle the objection | Hold rate at 50% or 75%, CTR, average watch time |
| CTA | Final 3 to 5 seconds | Convert retained attention into the next action | CVR or outbound CTR, depending on campaign objective |
The body is the longest part, but its job is not just to fill time. It has to place the offer before the halfway point and then defend it. The CTA at the end is the second ask, not the first introduction of what you sell.
Use the full hook-body-CTA breakdown if you need the anatomy. For testing, the table above is the operating model.
Why the offer belongs before the halfway point
If you are running a 30-second video, second 15 is not the midpoint of attention. It is the point by which a large share of viewers is already gone. Our 2026 fatigue benchmark puts the Meta Reels median hook rate at 28 percent and the CTR decline at 15 to 20 percent within the first two weeks. By the time the body starts, you are already writing for a minority.
Withholding the offer until after the halfway mark asks the remaining viewers to sit through the weakest part of the ad before they learn why they should care. A better sequence is: hook opens the problem or mechanism, body lands the offer and proof before the midpoint, and the CTA repeats the action. That way, the final 5 seconds are a reminder, not a reveal.
What varying only the hook tells you
Run two hooks against the same body, same CTA, same audience and same placement. If hook B lifts hook rate and early hold but downstream CTR does not move, the new hook is better at stopping the scroll but not better at attracting buyers. If hook B lifts hook rate and CTR, the new hook is pulling in people with intent.
For reference, we typically see TikTok in-feed hook rates of 20 to 30 percent and Meta feed hook rates of 15 to 25 percent on active accounts.
This is also where you catch a hook that overpromises. A hook might spike 3-second views and then show a sharp drop in the first third of the body. That is not a body problem. That is a hook problem, because the promise from the first 3 seconds did not match the next 10.
What varying only the body tells you
Keep the hook and CTA fixed, then run two different bodies. One body might open with the offer at second 4 and use a customer result. Another might spend the first 10 seconds on product features and delay the offer until second 18.
The body-only test isolates the middle. If the hold rate at 50 percent improves but CTR does not, your new body is holding attention but not changing intent. If CTR improves while hold rate stays flat, the body is more persuasive at the point of action even though retention did not move. If neither metric moves, the problem is usually not the body. It is the hook, the CTA, or the underlying argument.
This is why the offer belongs before the halfway point. A body that moves the offer from second 18 to second 5 will often lift hold rate simply because more people know what the ad is selling before they leave.
What changing only the CTA tells you
Fix the hook and body, then swap the final ask. A conversion objective might test "Start free" against "Book a 15-minute demo". A traffic objective might test "See plans" against "Get the report".
The CTA-only test tells you whether the retained audience has been given the right next step. If CVR moves, the new ask matches intent better. If nothing moves, the CTA is not the bottleneck. Do not spend another week trying to fix the last 3 seconds when the body lost the buyer at second 10.
Recombination vs a new argument
Recombination means assembling the same building blocks in new combinations: hook A with body C and CTA B, then hook B with body A and CTA D. You are testing expression, not the core claim. A new argument changes the problem you name, the promise you make, the proof you show, or the mechanism you offer.
If you have run 12 combinations of the same three parts and conversion is flat, you do not have a recombination problem. You have an argument problem. Go back and rewrite the body's offer or change the hook's problem statement. Modular video ads only help when the argument has life left in it.
This distinction matters because a lot of creative testing is just rearranging furniture. You can swap hooks and CTAs for weeks, but if the core promise is weak, the structure will not save you.
How Genyad fits this workflow
Genyad is built for isolating parts. You upload footage once, it transcribes and tags every clip, then each variation is a fresh script, shot selection, voiceover, caption set and export built from that library. That means a body-only test can pull different product shots and a rewritten middle voiceover while the hook and CTA stay fixed. It is not a re-cut of the same timeline.
The AI ad script generator writes scripts natively in the target language, not translated, so a German hook test is actually a German hook, not an English script with subtitles. Genyad does not have predicted performance scores and does not publish directly to Meta or TikTok. You run the variations where you already buy media.
If your bottleneck is an AI avatar speaking a synthetic script, Genyad is not the tool. If your bottleneck is turning footage you already own into many fresh hook, body and CTA combinations, it is.
Frequently asked questions
How long should a hook be?
Three seconds is the common benchmark on Meta and TikTok. Some hooks earn 5 seconds, but if the first 3 seconds do not stop the scroll, the extra 2 rarely fix it.
Does the offer always go before the halfway point?
For direct response, yes. The offer should appear early enough that the CTA at the end is a repeat, not a reveal. If you are running a brand spot with no direct action, the rule can bend, but most performance creative should not hold the offer past the midpoint.
How many variations should I run for a proper test?
Eight to twenty live variations is the range a typical active campaign needs, according to the fatigue benchmark we use. That is not all from one concept. Spread them across hooks, bodies and CTAs so you can isolate parts.
What does Genyad cost per variation?
One standard video ad variation costs 1 credit. The free plan gives you 5 variations and one AI-generated video up to 10 seconds. Starter is €29 for 15 credits, Growth is €99 for 65 credits, and Enterprise is custom. Credits never expire.
Can Genyad publish directly to Meta or TikTok?
No. Genyad exports in 9:16, 4:5, 1:1 and 16:9 at 1080p on self-serve plans. You download and upload to the ad platforms yourself.