Background. Decentralized participatory breeding approach facilitate the development of varieties better suited for a diversity of farmers' contexts. To enhance breeding efficiency, formal methodological settings are needed to include, at early stage of selection, particular local practices and farmers' preferences. Objective. Evaluate how location, farmers' profile and local cropping system could be taken into account to optimize selection of cowpea varieties in a decentralized context. Methodology: Twenty-two candidate multiline varieties of cowpea (Vigna unguiculata L. Walp) were developed from a biparental family of recombinant inbred lines. These varieties were subjected to selection under edaphic and climatic conditions of three locations and two cropping systems (intercropping versus sole cropping). Trials were carried out in collaboration with farmers' federations. Participatory evaluations were conducted over two years by the three groups of farmers defined by the following production goals: grains production as priority (“Grain priority profile”), fodder as priority (“Fodder priority profile”) and grain quality as priority (“Food processing priority profile”). Results. The statistical analysis supported significant effects of location and cropping system on the agronomic traits, with interactions effects involving the variety. A strong correlation was observed between the varietal choices of the two farmers' profiles which prioritized respectively grain and food processing. These farmers preferred varieties with higher grain yield and best grain quality. Farmers who prioritized fodder preferred more specific varieties, characterized by highest haulm yield. Implications. The clustering of farmers into specific profiles is an efficient method which allow expressing their diversified production goals through participatory evaluation. This led to more specific varietal choices for each of the profiles. Conclusion. This study set cropping systems and farmer pro