Testing creative does not mean launching videos until one works. It means changing one thing at a time and knowing in advance what you are measuring.

The minimum matrix

Two axes, not one:

  • Hook: how the first three seconds open. Situation, question, objection, result.
  • Angle: which argument carries the video. Price, time saved, quality, comparison with the alternative.

Three hooks across two angles gives six variants, and that is the minimum to learn anything. With fewer, you are watching noise.

How long to let a variant run

Until it has enough data to decide, not a fixed number of days. In practice, on small budgets, two to four days. Stopping after twelve hours because "it is not working" is the fastest way to throw away a variant that would have won.

When to kill one

When cost per acquisition is clearly worse than the others and it has already collected enough data. Not when the marketing team dislikes it. The creative the team hates and that sells is more common than people admit.

What to do with a winner

Do not run it until it burns out. As soon as it wins, produce two new variants from its hook, changing only the angle. That is how you build a series instead of restarting from zero each time.

The most frequent Italian mistake

Testing six videos from six different creators and concluding which creator is best. That is not the data: you changed two variables at once, person and content. To learn who works, you need two videos from the same creator with different hooks.

Keep reading

Find creators for your next test →

Frequently asked questions

What is the minimum number of variants?

Six: three hooks across two angles.

How long should a variant run?

Two to four days, until the data is enough.

When should I kill a variant?

When it has enough data and performs worse.

What comes after a winner?

Two new variants from the same hook.