Postpartum haemorrhage is the leading cause of maternal death in the developing country and yet is poorly diagnosed due to inaccurate measurement of blood loss following delivery. A study was carried out at Muhimbili National Hospital (MNH) Tanzania between 14^th October 2005 and 31^st January 2006 to determine the accuracy of visual estimation of blood loss (VEBL) in comparison to laboratory measurement of blood loss in diagnosis of primary postpartum haemorrhage (PPH). Four hundred twenty six pregnant women who were in active phase of labour were recruited and their venous blood was drawn for estimation of haematocrit before delivery and twelve hours thereafter. Active management of third stage of labour was conducted by giving 10 IU of oxytocin (intramuscularly or intravenously) and this was followed by visual estimation of blood loss. Proportion of patients who developed PPH was then determined by both methods. The mean duration of third stage of labour was 8.3 minutes and mean blood loss was164.9 mls. The prevalence of PPH was 8.9% and 16.2% by VEBL and changes in haematocrit respectively. Change in haematocrit in diagnosis of PPH was found to be more accurate, specific with high positive predictive values compared to VEBL.The need for additional uterotonics was 5.8% and the commonest labour complications associated with PPH were second degree tear, retained placenta and EUA for continued bleeding. In a busy labour ward like that in MNH service providers and medical students working in labour ward need to be taught regularly on how to estimate blood loss using simulated methods so as to increase their long term memory and accuracy in diagnosis of PPH hence provision of immediate intervention in case primary PPH occurs.