This article assesses the impact of the Northern Ghana Millennium Village Project. We estimate project effects on the Millennium Development Goals (MDGs) indicators using a difference-in-difference approach applied to matched villages and households using a sub-classification of the propensity score. The project improved some MDG indicators but, with few exceptions, impacts were small and core welfare indicators, such as monetary poverty, undernutrition and child mortality, remained unaffected. We found no spillover effects of the project to neighbouring areas and no displacements of development expenditure by local government and NGOs. We assessed the cost-effectiveness of the intervention and concluded that MVP did not produce the expected cost-saving synergies. We attribute the lack of impact to poor project design, redundancy of the interventions, and excessively high expectations.