BACKGROUND: Objective: This study aimed to use the Robson Ten Group Classification System (TGCS) to assess caesarean section (CS) rates and other outcomes in eight referral hospitals in Burkina Faso before the implementation of non-clinical interventions to reduce unnecessary CSs. DESIGN: This is a cross-sectional study. SETTING: We conducted a 9-month prospective observational study on women who gave birth at eight referral hospitals in Burkina Faso between 1 April 2020 and 31 December 2020. PARTICIPANTS: We analysed 24 643 women who gave birth at the eight participating hospitals during the study period. OUTCOMES MEASURES: We reported the relative size, CS rate and absolute contribution of each Robson group. These indicators were calculated for both referred and non-referred women. Oxytocin administration and stillbirth rates were calculated for women without previous CS and with a single fetus at cephalic presentation at term (groups 1-4). RESULTS: Overall, 24 643 women gave birth at the eight participating hospitals during the 9- month study period. The overall CS rate was 30.6%. Women in spontaneous labour with a single fetus in the cephalic presentation at term without previous CS (groups 1 and 3) had high CS rates (26.5% and 15%, respectively), low oxytocin use (7.9% and 6.5%, respectively), and high stillbirth rates (3.4% and 3.9%, respectively). These subgroups of women were major contributors to the overall CS rate. CONCLUSION: Our results indicate that, in referral hospitals in Burkina Faso, the CS practice for referred women in groups 1 and 3 of the TGCS should receive special attention. These results also reveal areas for clinical improvement to reduce primary CS, especially in nulliparous women. The use of the TGCS is important in low-income countries where low CS rates at the population level may conceal suboptimal labour management in healthcare facilities. TRIAL REGISTERATION NUMBER: ISRCTN67214403.