OBJECTIVE: To describe birth weight and postnatal weight gain in a contemporaneous population of babies born <32 weeks' gestation, using routinely captured electronic clinical data. DESIGN: Anonymised longitudinal weight data from 2006 to 2011. SETTING: National Health Service neonatal units in England. METHODS: Birth weight centiles were constructed using the LMS method, and longitudinal weight gain was summarised as mean growth curves for each week of gestation until discharge, using SITAR (Superimposition by Translation and Rotation) growth curve analysis. RESULTS: Data on 103 194 weights of 5009 babies born from 22-31 weeks' gestation were received from 40 neonatal units. At birth, girls weighed 6.6% (SE 0.4%) less than boys (p<0.0001). For babies born at 31 weeks' gestation, weight fell after birth by an average of 258 g, with the nadir on the 8th postnatal day. The rate of weight gain then increased to a maximum of 28.4 g/d or 16.0 g/kg/d after 3 weeks. Conversely for babies of 22 to 28 weeks' gestation, there was on average no weight loss after birth. At all gestations, babies tended to cross weight centiles downwards for at least 2 weeks. CONCLUSIONS: In very preterm infants, mean weight crosses centiles downwards by at least two centile channel widths. Postnatal weight loss is generally absent in those born before 29 weeks, but marked in those born later. Assigning an infant's target centile at birth is potentially harmful as it requires rapid weight gain and should only be done once weight gain has stabilised. The use of electronic data reflects contemporary medical management.