Through the work of French economist Esther Duflo, a particular method for public policy evaluation has come to prominence in recent years: randomized experiments. Popularized by their widespread use in poor countries since the early 2000s, they are inspired by randomized clinical trials in medicine. This article initially revisits the conditions of their emergence in the fight against poverty before focusing on J-PAL, the research center directed by Esther Duflo specializing in the approach. We then turn to the transition from theory to practice. The widespread pressure to publish in the academic field takes on a particular importance here, since randomized experiments are likely to have very concrete consequences on the populations concerned. The relatively distant relationship that researchers have with their field sites leads to a skewed account of local situations and development-related issues, which are related to a compartmentalization of disciplines that devalues qualitative methods and thus limits the effects of knowledge.