This Article identifies and critically assesses the crucial but troubled system for the coordination of international humanitarian assistance (the U.N. “Cluster Approach”). Regardless of whether the Cluster Approach actually “helps” in disaster response, it exercises substantial power over affected populations by assigning competences and leadership roles. The built-in mechanisms for controlling this power are unworkable, as they ultimately fail to resolve the tension between humanitarian organizations’ autonomy and the need for coordination. This Article identifies the emergence of an alternative model of accountability, based on mutual monitoring and “peer review”. Drawing on theories of “network” governance and “experimentalism”, this Article teases out the institutional and normative implications of such a model. In particular, the Article argues, a turn toward peer review would demand dramatic improvements in the inclusion of affected populations in the cluster system. This investigation may carry broader lessons for transnational networks and the study of accountability in global governance.