Assessment of open-ocean ecosystems relies on understanding ecosystem dynamics, and development of end-to-end ecosystem models represents an approach that addresses these challenges. These models incorporate the population structure and dynamics of marine organisms at all trophic levels. Satellite remote sensing of ocean colour and direct at-sea measurements provide information on the lower trophic levels of the models, and fisheries studies provide information on top predator species. However, these models suffer from a lack of observations for the so-called mid-trophic levels, which are poorly sampled by conventional methods. This restricts further development, and we argue that acoustic observations from a range of platforms (e.g. buoys, moorings) can be linked to the ecosystem models to provide much-needed information on these trophic levels. To achieve this, the models need to be tailored to incorporate the available acoustic data, and the link from acoustic backscatter to biologically relevant variables (biomass, carbon, etc.) needs attention. Methods to progress this issue are proposed, including the development of observation models and focal areas for ground truthing. To ensure full use of the potential of acoustic techniques, we argue that a systematic and long-term strategy incorporating the following elements is required: development of metadata standards and automated data analysis, inclusion of acoustic sensors in large-scale observatory programmes, improvement of observation-model links, and efficient sampling strategies. Finally, these elements should be tied together in an observation-modelling framework, coordinated by international organizations, to improve our understanding and quantification of open-ocean ecosystem dynamics.