Bed nets can prevent malaria, antiretroviral drugs extend life for HIV-infected people, and hand-washing reduces rates of diarrhea. Scientific studies have proven the worth of each of these. But there’s a world of difference between what works in a tightly monitored small study and what works in the far messier real world. As global health budgets have surged, increasing scrutiny has come to programs designed to roll out health interventions to communities or even entire countries. Building on the randomized, controlled study design used in clinical trials of experimental medicines, a field known as impact evaluation has developed a suite of methods to evaluate how well health programs designed for communities or entire countries actually work at preventing disease and saving lives. This story examines what evidence makes a compelling case that a large-scale intervention worked—or did not—and the surrounding debates about the strengths and limitations of impact evaluation.