Experimental design and metrics have become increasingly common in international assistance, as donor agencies have demanded rigorous forms of evaluation and monitoring. This article contributes to debates about the effects of an “evidence‐based turn” on interventions and recipients by exploring two questions: What constitutes evidence when it comes to everyday practices of aid at global scales? How are the goals of assistance affected? The article draws on collaborative research with an NGO and a group of social scientists who seek to improve child well‐being in El Salvador. It shows how evidence‐making was polysemic and costly, ultimately impacting the NGO’s planned intervention. This outcome, I argue, was not a matter of poor planning, but reflects structural, evidence‐making demands placed on global assistance at this historical conjuncture. Discussions among stakeholders about the trade‐offs between evidence‐making and assistance is a possible future route through the challenges described in this article.