1. Define the question
Write the question your batch explores: Does a close-up communicate the feature more clearly? Does a conversational opening make the explanation easier to follow?
Do this before choosing a variant count. More files do not automatically create a better experiment.
2. Keep the common parts stable
Confirm the product facts and base storyboard. Choose the dimensions to vary: hook, opening visual, presenter, voice or CTA.
A production batch can explore several dimensions, but a controlled comparison should isolate the variable you want to understand.
3. Name the differences
Label each version with what changed. Review the actual script, scene plan and rendered video to confirm the difference exists.
A draft is not a finished video. In ProductPost, variant planning and video rendering are separate steps.
4. Interpret results with context
Creative review tells you whether a video is accurate and coherent. Distribution data tells you how it performed in a particular placement and audience.
Do not label a draft a winner based on an invented prediction score. Use available real metrics, and record where and when the video ran.