This study assesses the performance of maternity care and its specific service components (preventive interventions in antenatal care, antenatal screening, referral, obstetric care) in Banke District, Nepal, using a set of structure, process, and output/outcome indicators. Data sources included health service documents in 14 first level health units and two hospitals, covering 1378 pregnancies and 1323 deliveries, structured observations, antenatal exit interviews (n = 136) and interviews with maternity users (n = 146). Coverage of antenatal care (28%) and skilled delivery care (16%) was low. In antenatal care, preventive interventions were only partially implemented (effective iron supplementation in 17% of users). On average one minute was spent on individual counselling per consultation. 41% of pregnancies were identified as high risk and 15% received referral advice, which was followed in only 32%. Hospital deliveries accounted for 9.8% of all deliveries. Hospital-based maternal mortality was 6.8/1000 births and the stillbirth rate 70/1000. High rates of stillbirth were observed in breech delivery (258/1000 births), caesarean section (143/1000) and twin delivery (133/1000). The risk of stillbirth was higher for rural women (RR 2.3; 95% CI 1.51-3.50) and appeared to be related to low socio-economic status. Emergency admissions were rare and accounted for 3.4% of hospital deliveries or only 0.4% of all expected deliveries. There was hardly any accumulation of high-risk pregnancies at hospital. The population-based rate of caesarean section was 1.1% (urban 2.3%, rural 0.2%). The estimated unmet obstetric need was high (82 cases or 61% of expected live-threatening maternal conditions did not receive appropriate intervention). The limited effectiveness of maternity care is the result of deficiencies of all service components. We propose a two-pronged approach by starting quality improvement of maternity care from both ends of maternity services: preventive interventions for all women and hospital-based obstetric care. Antenatal screening needs to be rationalized by reducing inflated risk catalogues that result in stereotypical and often rejected referral advice.