States often delegate compliance monitoring to international monitors. It is commomly assumed that these monitors will report accurately on the information they gather. However, the effectiveness of compliance monitors varies widely. Monitors may fail to collect information about non-compliance or even collude with non-compliant states and deliberately fail to report the information they gather. To explain this variation, we present a formal theory of how structural conditions and institutional designs lead to different levels of monitoring efficacy. We show that international institutions can improve monitoring by avoiding ruthless sanctions, and that intrinsically motivated monitors generally achieve better outcomes than neutral bureaucrats. Our theory contributes to the broad literature on international institutional design and treaty effectiveness.