Recent evolutions of the context of agricultural production in the world are calling for readily methods to help farmers in the design of innovative crop management systems (CMS) adapted to their specific farm constraints and soil-climate potential with a limited use of chemical input and energy. Methods for designing new CMS have long relied upon the interpretation of mono- or multi-factorial experiments, to propose technical recommendations and decision rules. Due to the high diversity of the techniques and the complexity of their combinations, CMS design have been relying increasingly on the use of simulation models (Bergez et al., 2010). Models allow to explore a wide range of biophysical and socioeconomic conditions and to end up with optimized combinations of techniques within the validity range of the models used. Nevertheless, the use of models is neither always possible nor desirable: their capacity to predict emerging properties from complex systems is limited; and there are situations, particularly in developing countries, where suitable models are not available, for example on certain crops. Usually, technical knowledge is available, but not formalized in academic standards or not organized in agronomic models. In these cases, we face the challenge of developing methods alternative to modeling in order to design innovative CMS, i.e. to mobilize expert knowledge from diverse sources and assess the combination of various technical changes by field experiments. Our contribution focuses on the review of a set of published experiences of CMS design without using agronomic model, to design complex combinations of techniques in order to produce innovative CMS adapted to particular environments and assessed with diverse performance criteria.