This study endorses the use of data envelopment analysis, which uses benefit-of-the-doubt weighting to evaluate the social, economic and overall performance of social enterprises. This methodology is especially useful for creating composite indicators based on multiple outputs expressed in different measurement units, and allows for enterprise-specific weighting of the different objectives. Applying this methodology on a unique longitudinal dataset of Flemish sheltered workshops suggests that social enterprises may face different types of mission drift. Further, our results show that top-performing social enterprises are more economically and socially efficient than low performers. These top performers also have a stronger economic orientation, which sheds new light on the balance between social and economic orientations in social enterprises.