Quantity of weigh-in-motion (WIM) data influences the results of bridge (and pavement) load modelling. However, no clear recommendations existed about how many vehicle records are needed for reliable prediction of load effects. Impact of quantity of WIM data on the results of bridge traffic load modelling was studied using three WIM dataset that contained almost one million heavy vehicle records. The main goals of the study were to investigate (a) variation of the bridge load modelling results as a function of WIM data sample size, and (b) how much data is needed for the results to converge. The datasets were quality-checked, which included visual validation of all heaviest vehicles. Vehicles with more than seven axles were eliminated to minimise the influence of random special transports. Convolution method was used as an efficient traffic load modelling tool to generate a multitude of different outcomes. Paper describes the approach taken and results from three locations with different traffic densities. Results for bending moments and shear forces are presented as a function of bridge span (5, 15, 25, 35 and 45 m) and size of data sample, which varies from 10 000 vehicles to the full dataset. Paper concludes with recommendations about the required number of WIM records for reliable bridge load effect modelling.