Linking experimentation, field data collection and modelling is essential in the design of alternative agricultural systems. Most alternative agricultural systems, such as conservation agriculture (CA), organic farming or ecologically intensive low-input systems, rely largely on biologically mediated processes and ecological-based support and regulation services. Since deterministic modelling of such systems typically fails to reproduce their intrinsic behaviour and complexity, experimentation remains indispensable in many cases. We show through examples that field measurements, experiments and modelling can support each other when one is diagnosing a specific problem or exploring viable alternatives as part of the design of, for example, disease-tolerant cropping systems, pest-suppressive landscapes or closely integrated crop-livestock farming systems. Expert knowledge and farmer perceptions may provide essential clues for validating models at scales transcending the field-plot scale. Instead of segregating modelling from field observations, experiments or practitioner knowledge, we propose a framework for their integration in the design of alternative agricultural systems.