<p>Philosophical discussions of how measurements improve social-scientific understanding are few and far between. Using the development of segregation indices as our primary case study, we fill this gap. We first distinguish four different kinds of measurement-based representations, and then show how each of these representations improves understanding, either by licensing novel modal inferences or improving explanations. Our discussion also highlights some underappreciated challenges for accounts of understanding that foreground coherence instead of modal inference or explanation.</p>