How many variants for a test, what to measure in the first 72 hours, and why the hook decides the budget.
The hook decides the budget
That is the part of the video commanding everything else.
The rest of the film only exists for those who got past the first three seconds, and that filter decides the cost of every result.
Three families of hooks to test in parallel, never one after the other.
The named problem: the annoying situation, shown before any mention of the product.
The result shown upfront: the end of the demonstration placed at the start.
And the contradiction: a sentence going against what the audience expects from an advert in that category.
How many variants for a test
Four to six per cycle, no more.
Below that, you cannot tell whether the gap comes from the video or from chance.
Above it, the budget scatters and no variant reaches the volume needed to be judged.
An economical production method: a single shoot, a single video body, and five different openings edited onto it. You test the variable that counts without paying for five shoots.
What to measure, in order
Four indicators, and the order matters as much as the figures.
Retention at three seconds, which judges the hook.
Retention at fifteen seconds, which judges the demonstration.
The click rate, which judges the promise.
And the cost per result, which judges everything else.
A rule of patience: cut nothing before a variant has spent enough to produce comparable results. A decision made too early regularly eliminates the video that would have won.
The Spanish detail few brands exploit
That is the point separating a local campaign from a translated one.
Three elements are worth testing separately.
The voice: a local accent is not replaced by a neutral voice, and the difference shows in retention.
The hour filmed: meals and the end of the day do not fall at the same hours as elsewhere, and a set at the wrong hour rings false.
And the vocabulary of the promise, which changes from one region to another far more than the product does.
One consequence for Latin America: the same video is not recycled as it stands. The price, the payment method, the vocabulary and sometimes the person have to be reworked, and the signed agreement must cover that additional territory.
What to deliver to launch a test
| Element | Useful quantity | Why |
|---|---|---|
| Hooks | Three per video | The first driver of results |
| Formats | Vertical plus square | Two placements |
| Subtitles | Burned in | Silent playback |
| A version with no text | One | Adaptation to another market |
What changes between two trials
One variable only. Changing the hook and the edit at once gives an unattributable result, and the budget spent teaches nothing.
What gets a video rejected at approval
An unverifiable claim, a before-and-after with no disclaimer, text promising a result, or a missing partnership disclosure when the brand has paid.
When to renew the creatives
Two signals, and a calendar.
The cost per result rises three days in a row while nothing has changed in the targeting.
The same person sees the video too often, and comments become comments about the advert, not about the product.
The calendar: prepare the next batch before you need it, because a shoot takes two to three weeks between brief and delivery. A campaign stopping for lack of fresh creatives costs more than the shoot nobody wanted to order.
One last obligation, valid for all these variants: the advertising disclosure is still owed in each one, including in the short versions, and its exact forms are checked with the competent authorities.
Sources
Checked on 4 September 2026. Thresholds and rates change: where this guide and the official source disagree, the official source is right.
