Convective heat transfer from exterior and interior surfaces of the building envelope to the surrounding air is usually modelled by convective heat transfer coefficients, which depend on the wind speed and air flow pattern near the surface in case of forced convection. These coefficients are required for energy performance analysis of buildings and to asses the hygrothermal behaviour of building envelopes. The convective heat transfer coefficients at exterior surfaces are mostly described by empirical correlations as a function of the wind speed measured at a meteorological station or at some distance from the building surface. These correlations were obtained by laboratory experiments on flat plates or by full-scale experiments on buildings. While the former somehow lack physical similarity, the latter only provide data at a limited number of facade locations. In general, a uniformly distributed convective heat transfer coefficient over the entire facade is assumed. Recently, Computational Fluid Dynamics (CFD) has also been used to obtain the convective heat transfer coefficient distribution at building facades. In building aerodynamics, steady Reynolds Averaged Navier-Stokes (RANS) simulations are commonly used to model air flow combined with wall functions to model the boundary layer because of the low computational cost. With this technique however, the boundary layer is modelled partially and the unsteady character of the air flow can not be captured. This simplified air flow modelling will consequently affect the accuracy of the predicted heat transfer coefficients. In this paper, CFD is used to asses the exterior convective heat transfer coefficient distribution on the facades of a cubic building. Wall-function modelling is not used but instead, the boundary layer is fully resolved which requires an increased amount of computational cells in the near-wall region. Steady RANS simulations are performed as well as simulations with a more advanced turbulence modelling technique, namely the Detached Eddy Simulation model, which is inherently unsteady. The convective heat transfer coefficient is found to vary significantly over the different facades and its distribution is closely related to the flow field around the cubic building. The heat transfer coefficients of steady and unsteady simulations are compared and, apart from the windward facade, significant discrepancies are found. This is related to the unsteady nature of the separation bubbles at the roof and side walls and the periodic vortex shedding phenomenon in the wake of the building, which can not be captured with steady simulations. Correlations of the convective heat transfer coefficient with the wind speed measured at a meteorological station are provided for various locations on each facade and are compared with existing empirical correlations.