Objective: To estimate the proportion of caesarean sections (CS) not meeting audit criteria for prolonged labour. Design: Cross-sectional. Setting: Five urban maternity units in Dar es Salaam, Tanzania. Population: Women giving birth by CS with an indication of prolonged labour, from October 1 st, 2021 to August 31 st, 2022. Exclusion criteria: referral to the study sites because of prolonged labour or cervical dilatation >6 cm upon admission; non-cephalic presentation; multiple pregnancy; intrauterine fetal death; failed induction; previous CS; or other reasons for CS. Methods: Criterion-based audit of CS case files with an indication of prolonged labour. Main Outcome Measure : CSs in women with uncomplicated labour progress. Results: Overall CS rate was 32% (2949/9364) and 746/1517 (47.9%) of first-time CSs were performed because of prolonged labour. Out of these, 456 met inclusion criteria and 243/456 (53.3%) CSs were in uncomplicated labour: 1) women not being given a trial of labour (78/243, 32.1%); 2) women in first stage of active labour not crossing the partograph action line (145/243, 59.7%); and 3) women in second stage less than 1 hour (20/243 8.2%). Conclusion: Almost half of CS in the unscarred uterus were because of prolonged labour and many did not meet audit criteria for prolonged labour. Crowded hospitals and inadequate monitoring may have prompted defensive decision-making. Unconducive labour wards may, therefore, indirectly drive the CS epidemic while clinical guidelines for CS decision-making remain scarce.