With increasing pressure on farmers to improve the performance of their cropping systems, there is a growing need to design cropping systems that respond concurrently to environmental, agronomic and socioeconomic constraints. However, the trade-offs between ecosystem services, including provisioning services, can vary considerably from plot to plot. Using a typology of agricultural practices to adapt a conceptual model of the cropping system can provide an instrument to support the design of cropping systems that take into account the diversity of environmental and socioeconomic conditions and trade-offs within a study site. This method was tested to design coffee-based agroforestry systems mitigating soil erosion in central Costa Rica, a case study with a high-value crop in a complex relationship to its biophysical environment. A generic plot-scale conceptual model representing the effect of agricultural practices and environmental factors was designed, with erosion reduction, coffee production and gross margin as the outputs. Quantitative data on agricultural practices and costs were then collected over two years on a sample of plots in an 18 km(2) watershed upstream of a hydroelectric dam. A typology of these plots was built based on agricultural management practices; the resulting groups were further characterized by socioeconomic and environmental variables. The critical variables from each group of plots were used to adapt the conceptual model to the groups from the typology. The four groups found were (1) low-intensity management; (2) intensive management; (3) shaded agroecosystem, and (4) intensive agrochemical management. The conceptual model helped analyze the key processes and trade-offs for each group and helped make recommendations of adapted erosion control practices. The model showed that for some groups, less time-consuming erosion control actions not impacting coffee production might be more suitable, such as drainage canals, terraces, and vegetative barriers. In contrast, other groups had better socioeconomic or environmental conditions that opened the possibility of using shade trees or manual weed control (as opposed to herbicide use) to control erosion. This method finds its application in the plot-scale design and prototyping of agricultural systems that better respond to specific constraints, and can provide more relevant basis for discussion with farmers in participative methods. It also presents the advantage of requiring little data acquisition, although it can be further developed through integrating numerical relationships for quantitative modeling. (C) 2013 Elsevier Ltd. All rights reserved.