A test screening runs your cut past five simulated audience personas and reports predicted engagement, drop-off points and scenes likely to confuse.
This is a structural heuristic, and it is labelled as one. It reads pacing, scene length, dialogue density and tone-map shape — it is not a real audience, and it does not claim to be. It is useful for catching a fourteen-minute second act with no turn in it; it will not tell you whether your film is good.
A/B testing compares two thumbnails, titles or trailers, measured on click-through, watch time and completion, with a confidence gate so you are not told a 3% difference on 40 views is a result.
Both are replaced by real Theater analytics once a title has an audience.