OBJECTIVE Hepato-pancreato-biliary (HPB) surgeries are one of the most challenging and complex procedures. Intraoperative frozen section (IFS) diagnosis plays a pivotal role in management decisions. Comprehensive large cohort studies evaluating utility of IFS in HPB malignancies are lacking. This study aimed to evaluate the accuracy of frozen section analysis and to analyse discrepancies and impact of IFS on the surgical decisions. METHODS This was a retrospective study of IFS received for the HPB specimens between years 2009 and 2021. The results were compared to the permanent sections to evaluate diagnostic accuracy, sensitivity and specificity. Indications, disagreements and impact on the surgical management were analysed. RESULTS A total of 1008 specimens were evaluated: bile duct margin (279; 27.7%), gallbladder (203; 20.1%), liver lesions (125 cases; 12.4%), lymph nodes (147; 14.6%), pancreatic margin (120; 11.9%) and deposits (134; 13.3%). IFS were diagnosed as negative for malignancy (805; 79.9%), positive for dysplasia (8; 0.8%), suspicious for malignancy (6; 0.6%) and positive for malignancy (189; 18.8%). The overall diagnostic accuracy was 98.4%, and the discordant rate was 1.6%. The sensitivity, specificity, positive predictive value and negative predictive value were 94.7%, 99.4%, 97.5% and 98.6% respectively. The most important reason of discordant results was technical, followed by interpretational and sampling errors. CONCLUSIONS The study demonstrates high diagnostic accuracy (98.4%) of IFS in a large dataset of HPB specimens. This comprehensive analysis apprises of the indications, errors and the impact of IFS diagnosis on subsequent HPB surgical management.