Despite a steady growth in coral-reef monitoring efforts, the application of the monitoring results to decisionmaking often remains limited, because questions that can be answered are frequently posed after monitoring commences, rather than having the questions define the data to be gathered. We review how hierarchical, question-driven frameworks can improve monitoring designs and how added attention to high-population-variance structures play a central role in this process. Stratification is necessary to avoid the high variance and low power caused by sampling across coral-reef habitats. Yet, knowing when and where to introduce stratification into sampling designs requires information on the environmental and biological processes that drive species abundance patterns. Using case studies, we review some limitations of approaches that back-calculate the effort required to attain desirable statistical power and highlight some approaches to better account for the heterogeneous nature of coral-reef assemblages in monitoring designs.