Long-term research and monitoring can provide important ecological insights and are crucial for the improved management of ecosystems and natural resources. However, many long-term research and monitoring programs are either ineffective or fail completely owing to poor planning and/or lack of focus. Here we propose the paradigm of adaptive monitoring, which aims to resolve many of the problems that have undermined previous attempts to establish long-term research and monitoring. This paradigm is driven by tractable questions, rigorous statistical design at the outset, a conceptual model of the ecosystem or other entity being examined and a human need to know about ecosystem change. An adaptive monitoring framework enables monitoring programs to evolve iteratively as new information emerges and research questions change.