Data Envelopment Analysis (DEA) was used to measure efficiency of residential mental health facilities. The sample consisted of 50 half-way houses, 8 nursing homes, and 32 sheltered homes. In total, 68 facilities belonged to the public sector and were 22 supervised by private non-profit organizations. Variables chosen to characterize production were: structure size (m2), staff, salaries and operational costs, and the output measure was patient numbers. An input oriented DEA model, allowing for variable returns to scale, was applied and units were ranked according to a benchmarking approach. Mean efficiency, for the whole sample, was 73.2% and 18 best practice units were found, on average, 33.1% over-efficient. The other 72 were under-performing, with 54 appearing more than 20% inefficient. The mean efficiency scores for public and private non-profit units were 68.8 and 86.6%, respectively, and significantly different (p < 0.001). Results suggest that efficiency improvements are possible with better use of resources but more research employing various data sets is required.