The study objective was to evaluate three methods of Salmonella enterica prevalence estimation in swine herds (faecal culture, culture of abattoir-collected samples, and serum ELISA). From each of six swine herds, we necropsied approximately 100 finishing pigs (>70 kg); one-half on farm and the other half at the abattoir, after transport and approximately 2·5 h holding. We collected the same samples for S. enterica culture at both locations (1 g faecal, 10 g caecal contents, ileocaecal lymph nodes, superficial inguinal lymph nodes, 25 g of gluteal muscle for serum ELISA). On farm, the 1 g faecal sample only detected 13·3% (2/15) of all positive pigs necropsied on farm. However, with abattoir and on-farm results combined, the faecal sample detected 57·4% (74/129) of positive pigs. Abattoir-collected samples provided prevalence estimates much higher than on-farm collected samples (39·9 vs. 5·3%; P<0·001). This study shows that faecal samples have a low sensitivity for detecting infected pigs and that abattoir-collected samples overestimate the on-farm S. enterica prevalence. For most herds, serology overestimated the on-farm culture prevalence.