Objective: To assess the effect of serial growth scans and optimised fetal weight limits on the risk of stillbirth in low and high-risk maternity populations. Design: Retrospective cohort study Setting: United Kingdom 2015-2020 Population: 1,572,817 singleton pregnancies cared for in maternity units that have implemented the Growth Assessment Protocol (GAP). Methods: Analysis of fully anonymised, prospectively recorded core data from 132 National Health Service institutions. Stillbirth rate and relative risk was assessed according to risk status assigned at booking. We constructed receiver operator curves (ROC) and determined area under the curve (AUC) and optimal centile points using Youden’s Index. Main Outcome: Rate of stillbirth from 24 weeks gestation. Results: The overall cohort included 6569 stillbirths (rate per thousand: 4.18). The rate was higher in pregnancies that had been designated high risk (6.23) than low risk (3.61; RR 1.7, CI 1.6-1.8). High risk pregnancies that did not receive monitoring by serial ultrasound had a stillbirth rate that was more than twice as high than those that did get serial scans as per protocol (11.94 vs 5.64). The optimal centile point for predicting stillbirth was 11.3 for the overall cohort (sensitivity 36.8, specificity 84.1 and AUC 62.3%), 15.2 for low risk pregnancies (sensitivity 35.1, specificity 84.6, AUC 61.3%) and 2.7 for high risk pregnancies (sensitivity 34.2, specificity 88.9, AUC 61.2% ). Conclusions: Serial third trimester growth scans can halve stillbirth risk in pregnancies designated high risk. Optimal fetal size limits for antenatal surveillance are specific to the risk status of pregnancy.