The reflexive test
Apply the system's own standard to the sentences the company writes.
Throughout the product, a reading is required to state what it is based on and how firm it is — a finding with a thin sample is labelled as such, a benchmark states when it was built, an unconfirmed excess is called a candidate rather than a result. Those are not decorative habits; they exist because a claim without its confidence attached will be read at full strength. Product statements are read the same way by the same people. Applying a weaker standard to them than to the machinery would be a small hypocrisy with a real cost, since the buyer has no way to tell which sentences were held to which standard.
An ordinal scale for reading product statements, not a measurement. Anything above the lowest rung should be assumed to be capable of not arriving.

