The confidence meter: self-conviction = 0.1/10 confidence. A competitor having the feature = low confidence. Actual user tests and experiments = medium to high confidence. Most teams stop way too early and call it validated.