Development practitioners still lack a critical mass of empirical evidence which can help identify the set of interventions that are more likely to work, and inform the design and implementation of feasible reforms. This paper contributes to fill this gap by looking at the case of the 'Sierra Leone Pay and Performance Project', a World Bank-supported initiative to reform the civil service. It analyzes the functional problems characterizing the civil service and discusses what factors account for the observed dysfunctions. The central argument is that the current dysfunctions might be difficult to reverse as they define a sub-optimal equilibrium which serves political purposes (dysfunctions by design). However, politics is not all that matters. This equilibrium is further reinforced by systemic dysfunctions that may not be the consequence of any strategic design or the outcome of elite preferences (dysfunctions by default). This is where there is scope for change, even in the short run. The authors conclude that the chances of successful civil service reforms are likely to be maximized if reform initiatives support modest and incremental changes that work with the grain of existing incentives and are consistent with government preferences. The Sierra Leone Pay and Performance Project aims to do so by adopting a limited and targeted focus on pay reform, performance management and recruitment and staffing. In addition, the use of the results-based lending instrument is expected to help mitigate the current dysfunctions by aligning the incentives of the various players and, in this way, create the conditions for greater coordination across government agencies. Although the suggested approach is not without risks, recent dynamics suggest that the chances of success are greater today than in the past.