Monitoring progress in reducing maternal and perinatal mortality requires suitable indicators. The density of emergency obstetric care (EmOC) facilities has been proposed as a potentially useful indicator, but different UN documents make inconsistent recommendations, and its current formulation is not associated with maternal mortality. We compiled recently published indicator benchmarks and distinguished three sources of inconsistency: (i) use of different denominator metrics (per birth and per population), (ii) different assumptions on need for EmOC and for EmOC facilities and (iii) failure to specify facility capacity (birth load). The UN guidelines and handbook require fewer EmOC facilities than the World Health Report 2005 and do not specify capacity for deliveries or staffing levels. We recommend (i) always using births as the denominator for EmOC facility density, (ii) clearly stating assumptions on the proportion of deliveries needing basic and comprehensive emergency obstetric care and the desired proportion of deliveries in EmOC facilities and (iii) specifying facility capacity and staffing and adapting benchmarks for settings with different population density to ensure geographical accessibility.