T he growing use of randomized field experiments to evaluate public policies has been one of the most prominent trends in development economics in the past 15 years. These experiments have advanced our understanding within a broad range of topics including education, health, governance, finance (credit, savings, insurance), and social protection programs, as summarized in Duflo and Banerjee (2017). In this paper, we argue that experimental evaluations could have a greater impact on policy if more of them were (literally) bigger. We believe this for two reasons.First, large-scale evaluations can directly inform large-scale spending decisions. Governments (regrettably) often do not follow a process of testing prototypes and scaling up those that work. On the contrary, they often roll out new programs representing millions (or billions!) of dollars of expenditure with little evidence to indicate whether they will work. Randomizing these rollouts can generate direct evidence on policy questions that are inarguably of interest-after all, such programs are already heavily funded. Working with governments to evaluate these programs as they are being deployed, and before political constituencies have calcified around them, thus represents a tremendous research opportunity with immediate policy applications.