When performing an urban flood risk analysis, it is often difficult to take individual buildings into account: doing so requires the availability of a high resolution 2D hydrodynamic model for the preparation of flood maps and detailed land use maps for the preparation of flood damage maps. As a consequence, a simplified approach is often required, involving the use of low resolution models and simplified land use maps. This study aims at evaluating the impact of such simplifications on the flood risk by means of a case study: the flooding of the city of Antwerp (Belgium) caused by wave overtopping of the flood defenses along the river Scheldt. Two methods for computing flood maps were combined with two methods for computing damage maps, yielding four different methods for computing urban flood risk. The results obtained with the four methods differ significantly. The flood risk predicted by a combination of the detailed approaches was found to be less than 30% of the flood risk predicted by a combination of the simplified approaches. From this study, we can conclude that the procedures used for dealing with the presence of buildings can be a significant source of uncertainty in urban flood risk analysis.