The method
Claim classes
Every statement the narration makes falls in one class. Name the class before choosing a graphic; the class limits what the graphic may do.
| Class | Example | What may be shown |
|---|---|---|
| Measured | “Just Blue eats 4.2 stops on a B10X.” | The number, its unit, and where it came from. A spec callout or evidence card. |
| Observed | “The edge has worn through after six months.” | The footage or photo of the thing. A label at most. Never a graphic standing in for the shot. |
| Opinion | “I would not buy it again at this price.” | The words, in the creator’s voice, or nothing. Never a rating, score, star count, or checkmark. |
| Context | “I backed it on Kickstarter in 2024.” | A timeline mark, a date chip, a receipt-style card with real details. |
| Comparison | “Cheaper than the case I replaced.” | A two-column card with both sides named. Never a winner badge. |
The useful property of this table is that it is checkable. You can take any frame of a finished video, find the sentence underneath it, name the class, and see whether the graphic exceeded its licence. That check does not require taste. It requires reading.
The one that catches people
Observed is where most work goes wrong. The claim is “the edge has worn through,” the evidence is a shot of the worn edge, and the temptation is to make a diagram of wear because the real shot is unglamorous. The moment a graphic stands in for footage that exists, the video is asserting something it did not record.
If the shot does not exist, the honest options are to go shoot it or to cut the claim. Illustrating it is not one of them.
Related: the certainty rule governs how much confidence a graphic may carry; anti-slop lists what is banned outright.
Part of the method, v1.0. Revisions are recorded in what changed.