Increasing irrigation efficiency in irrigation schemes has always been high on the agenda of policy makers. Despite quite some 'social' experiments, whereby large portions of management were carried over to the farmer level, results often remained disappointing. This paper explores why this came about for a case study in the Office du Niger irrigation scheme in Mali. Since Irrigation Management Transfer, farmers are responsible for the tertiary level, but collective action for water management remains below expectations. This paper applies an analytical framework based on the conceptual model developed by van Noordwijk et al. (Conserv Ecol 5:21, 2002) and proceeding in two steps. Firstly, an interview-based stakeholder analysis assesses the objectives and mental models of farmers and the central management regarding water management. It appears from the interviews that the central management wants to increase irrigation efficiency through fully-fledged collective action, whereas farmers value the latter only when it favours easy irrigation. Secondly, the relation between collective action and performance was tested through a field study on a sample of 36 tertiary blocks. Results indicate that only collective action at the inlet of the tertiary canal, currently implemented on less than a third of the tertiary blocks, increases irrigation efficiency (with 14%). Collective action for water allocation is implemented within almost three quarters of tertiary blocks and effectively reduces irrigation problems. However, if they lack the necessary social capital, not all farmers can establish collective action when needed. Based on this analysis, the paper proposes a mix of incentives and measures to resolve the conflict between farmers and the central management to their mutual benefit.