OBJECTIVE: To assess the internal consistency of the Hadlock normal range for estimated fetal weight (EFW) and its clinical implications. DESIGN: Methodological analysis and retrospective observational audit. SETTING: Tertiary maternity hospital in Oxford, UK. POPULATION: Women with a non-anomalous singleton pregnancy. METHODS: Reanalysis of the original Hadlock reference EFW chart using functional data analysis (FDA) and penalised functional regression (PFFR-GAM), and systematic bias assessment across gestation. Retrospective EFW centile analysis in a population of pregnancies undergoing a growth scan between 35 + 0 and 36 + 6 weeks, calculating the EFW centile using both the published reference chart (chart-derived centile) and the reported regression equation (equation-derived centile). OUTCOME MEASURES: Agreement between chart-derived and equation-derived centiles and the corresponding differences in classification of fetuses below the 3rd and 10th centiles. RESULTS: There is a statistically significant disparity between the chart-derived and equation-derived centiles, most pronounced in the 3rd and 10th centiles where clinical implications are critical. The 3rd and 10th chart-derived centiles approximately correspond to the 1st and 6th equation-derived centiles. Conversely, the equation-derived 10th centile corresponds to the chart-derived 17th centile. In a population of 21 874 pregnancies undergoing a growth scan between 35 + 0 and 36 + 6 weeks, the chart classified 4.2% as below the 10th centile, compared with 7.2% using the equation, with an absolute difference of 3.0% (95% CI 2.8%-3.3%; McNemar's test; p < 0.001). Also, the chart classified 0.8% under the 3rd centile, compared with 2.3% using the equation, with an absolute difference of 1.5% (95% CI 1.3%-1.7%; McNemar's test; p < 0.001). CONCLUSION: Systematic disparity exists between the published chart- and equation-derived centiles in the original Hadlock growth reference. The equation-derived centiles are closer to the expected nominal proportions of our population, whereas the chart-derived centiles significantly underestimate the incidence of small for gestational age.