Non-judicious oxygen use in preterm infants is associated with increased risk of retinopathy of prematurity, bronchopulmonary dysplasia and longer hospital stay. Despite established guidelines on oxygen therapy, compliance with the best oxygen practices remains suboptimal. Excessive use of oxygen also consumes a large proportion of the annual maintenance budget of special newborn care units (SNCUs) in the districts. In this project, we aimed to reduce the oxygen consumption in the SNCU at Sehore, Madhya Pradesh, India from eight to four cylinders per day, by rationalising the indications, monitoring and method of oxygen delivery.We tested two sets of interventions using the Plan–Do–Study–Act (PDSA) approach. The first intervention was the introduction of a written ‘oxygen policy’ regarding indications of starting/stopping oxygen and the use of saturation targets. The second was using short binasal infant prongs (at 0.5–1 L/min), instead of oxygen hoods as the primary method of oxygen delivery in spontaneously breathing neonates requiring oxygen. In the first PDSA cycle, we assessed the feasibility of the intervention in a small set (n=30) of neonates and later scaled up to all eligible neonates in the second phase.We observed a significant reduction in oxygen consumption (from median (IQR) 8 (7–8) to 3 (3–4) cylinders per day) that can lead to a direct saving of 590 000 Indian rupees (US$9000) per year. There was a significant reduction in the number of neonates on oxygen support on a given day. We did not observe any increase in mortality or nasal injury. The change was sustained for the next 8 months.We conclude that by having a contextual oxygen policy and using nasal prongs instead of oxygen hoods as the preferred delivery method, we can achieve a sustainable reduction in oxygen consumption.