Modeling is increasingly used to explore a variety of technical innovations and evaluate ex-ante their impact on the functioning and performances of farming systems. These models usually include agronomic variables characterizing the management processes of cropping systems and their impact on expected yields. Crop models seem a powerful option to represent these processes but (i) they are often restricted to some parts of cropping systems; (ii) they usually require a large amount of input data that is not always available at plot and farm level; (iii) their domain of validity is restrictive in contrast to the farm diversity in a given region. Thus they are seldom used in decision support tools. In this paper we investigate various alternatives to characterize agronomic variables in tools aimed at supporting farmers' or food processors' tactical and strategic decisions.