Road traffic loading is a highly variable process with a wide range of complex variables to consider. Consequently, various simplifying assumptions are required to efficiently model heavy traffic events for bridge safety verification, which reduce the accuracy of the final result. For existing bridges, direct monitoring of the structure with accurate measurement systems can remove many of these uncertainties. This paper investigates how close simulations really are to the action effects arriving in structural elements as measured directly by sensors. The research uses an experimental setup in Switzerland consisting of an instrumented prestressed concrete highway bridge and a pavement WIM system close to the bridge. Rebars in the deck slab of the bridge are instrumented with strain gauges to measure the real traffic action effects. The difference between measurements and simulated traffic results is investigated. There is considerable variation between the results obtained from the two methods for the monitoring period of over 250 days. The difference in extreme values as predicted by extreme value theory is also studied and compared with the associated Eurocode load model results. This work highlights some of the limitations associated with current bridge modelling and makes the case for direct measurement of the structure in cases where high accuracy local information is required.