You made one video, it did well, and you know nothing. That is not a criticism of the video.
The variance is inside the creator, not between creators
What that changes about the number of videos ordered
When one person delivers five videos, the performance gap between their best and their worst is generally larger than the gap between two different creators. In other words, looking for the right creator is less profitable than giving the same one several chances. Ordering five videos from one person therefore produces more information than one video ordered from five people, for an identical budget.
The bottleneck moves to review
As soon as production increases, it is no longer shooting that limits, it is approval. Fifteen videos a month passing through three reviewers produce an immediate jam, and the videos arrive after their useful window. Volume is therefore not decided at the level of the production budget, but at the level of the number of approvals your organisation can absorb.
Vary only one thing
Fix the offer, move the opening
The product, the offer, the landing page and the duration stay identical; only the opening changes. That is the only way to attribute a result to a cause. A series where the creator, the hook and the edit all change at once produces no usable information, only an impression.
Knowing when a video is finished
Seven signals indicating you should stop touching it.
Retention at three seconds is stable over two consecutive days.
Cost per acquisition no longer moves despite the changes.
Frequency rises, which signals fatigue rather than a flaw in the video.
Comments become repetitive, a sign the message has landed.
The corrections requested concern preferences, not errors.
The video has reached the spend volume the test planned.
And you can no longer say which version is the right one, which means none of them is any more.
What to do with a winner
Seven actions, in this order.
Writing down precisely what won, in one sentence.
Buying the rights for a longer period, before the creator notices.
Ordering three variants of the same hook from the same creator.
Handing the same hook to three other people, dealt with above.
Adapting it into different formats, short, long, square.
Watching frequency and preparing the successor before fatigue.
And documenting everything, because the same idea will be proposed again in six months by somebody who does not know.
The review bottleneck, in practice
Six measures that unblock volume.
Two reviewers maximum, with a single right of veto.
A fixed approval slot, twice a week.
A deadline beyond which the video is deemed accepted.
A written list of what is approved and what is not up for discussion.
A feedback template where everybody writes what they would remove rather than what they would add.
And a simple rule: no changes after release, except for compliance.
How many to order, and from whom
Six sizing markers.
Five videos from one creator for a first test, rather than one from each.
Three different creators only once the hook is identified.
An identical budget per video, decided before launch.
A thirty-day measurement window, written in advance.
A single objective, the same for every video in the series.
And a production reserve planned for what follows, since a successful test immediately creates a need.
Why the second campaign disappoints
Six explanations, almost always the same.
The most receptive audience was already reached by the first.
Frequency rose without being watched.
The season changed, and the comparison is no longer valid.
The videos were improved in between, which made them more conventional.
The budget was raised all at once, which widened the audience beyond the target.
And the new series varied three elements at a time, which makes the result unreadable.
What you test when producing in series
| The variable | What it reveals | Number of versions |
|---|---|---|
| The hook | What stops the scroll | Five |
| The face | Who fits the audience | Three |
| The core argument | What decides the purchase | Two |
What makes a test readable
One variable at a time, an identical budget per version, and enough time to exit the learning phase. Without those three, the resulting ranking means nothing.
What costs less than people think
Multiplying hooks on one video body. Five openings cut onto the same shoot cost an hour of editing, whereas five shoots cost five days.
What to write in the contract for volume testing
Five clauses specific to testing.
The number of videos and a staggered delivery schedule.
The right to order variants of the same hook at an agreed rate.
The buyout of rights for a longer period, priced in advance.
Permission to hand the same hook to other creators, with what is owed to the original author.
And retention of the rushes for a defined period, indispensable for re-editing a winner.
One last marker: in the United Kingdom, a brand testing in volume is not looking for the best video, it is looking for the sentence that works. Once that sentence is identified, it is worth more than any file, because it can be shot again, by somebody else, in another setting, for two years.
