There is a strong tendency for evidence-based approaches to social practices to view these practices as imperfect devices for delivering social services. Practices are regarded as in need of repair by evaluation (and research) that can deliver the necessary science-based solution to the problems of practice. This article presents a different view of practices as material and linguistic events in which activities and relationships are constituted and unfold in interaction and in which people change and develop, and it argues for restoring this view of practice to evaluation. The article discusses two different ways in which notions of "evidence based," "practice," and "evaluation" are related and suggests what a genuinely practice-oriented approach to evaluation entails.