In this paper, I investigate the scalar semantics of evaluative adjective in general, and of good in particular. Lassiter (2017) has argued that good, when taking propositions as arguments, has an interval scale. I argue that there’s evidence in support of the view that good, when taking individuals as argument, has a scale that is stronger than interval, but weaker than ratio. In particular, I propose that individual-level good has a “round” ratio scale, which allows a broader set of ratio transformations than standard ratio scales. This conclusion is consistent with the fact that good admits round ratio modifiers (twice as good), but eschews precise ones (# 1.38x as good). An important consequence of this view is that the scales of individual and propositional-level good are severed.