The design of sustainable cropping and livestock systems has become a research priority. Model based design approaches have been criticized e.g. for being too complex to stimulate farmers’ learning and lead to effective innovation. This article presents an approach combining plot‐scale diagnosis and farm‐scale simulation tailored to support the design of novel grassland‐based beef systems. Because of its intelligibility and transparency, diagnosis is expected to constitute a suitable entry point for the subsequent model‐based design. Diagnosis determines the way in which the timing and intensity of farmer’s practices on a grassland plot are suited to the productive potential of this grassland or its assigned function. Based on these elements, it suggests adjustments to grassland use enabling novel systems to be devised. Simulations of current and novel systems provide daily variation of standing herbage, forage stocks and animal performance for different weather patterns. The simulated behaviour of current and novel systems can then be compared. The approach is applied to two grassland‐based beef systems in the French Pyrenees. Simulations reveal that improvement of forage self‐sufficiency pinpointed by the plot‐scale diagnosis was impractical at the farm scale due to weather and management constraints. Compared with diagnosis, simulations contributed to deeper learning of both scientists and farmers because of their level of integration and dynamic representation. Nevertheless, and as expected, diagnosis constituted a key stage in conditioning farmers to learning during the subsequent design and farm‐scale model‐based evaluation.