Cotton cultivation has expanded throughout West African savannahs following relatively standardised growing recommendations. Occurring institutional changes result in more diverse growing practices, less favourable zones under cropping and an overall diversification of cotton growing conditions. Current recommendations are then showing limitations and we try to develop a methodology for conceiving entire cropping systems that could be more adapted to most usual combinations of major constraints. This methodology is based upon expertise, knowledge or experience, whether detained by researchers, extension agents, formers or users. It involves three successive steps or "rendez-vous" devoted to: (i) The diagnosis of major constraints (identification and typology), (ii) The elaboration of the cropping prototype adapted to a chosen set of constraints and (iii) the evaluation and adaptation of the initial prototype after field testing. We present a case where cotton is grown with very limited water available, either because rains are scarce or because of late sowing. The prototypes formed by the experts have been tested in ten locations scattered between Benin, Cameroon and Mali. First results show that the should be specifically adjusted to local constrain before entering a second year of testing, Because it aims at answering a quite complex and very applied question, this approach provides an opportunity for mixing researchers and other actors of the sector and it also draws a good frame for planning research activities within a multidisciplinary group or researchers.