This paper presents a comparison between CFD-simulations and measurements of the temperature stratification in a mixing box of an air-handling unit. We have used data from field measurements during a period of over a year for different outside temperatures. We performed two-dimensional CFD-simulations for four different outside temperatures with commercially available software. The measurements as well as the simulations show that the temperature difference between the upper part and the lower part of the duct downstream of the mixing box is considerable. It increases, as the outside temperature decreases. However, the discrepancies between the measurements and the simulations are large. The reasons for this are uncertain boundary conditions and modelling errors leading to an inaccurate simulation result. The stratification downstream of the mixing box implies large sensor errors and the use of the mixed air temperature for control and fault detection must therefore be questioned. Averaging sensors, which take a mean value over the duct section, can be used but do not consider differences in velocities and are therefore not accurate either. In order to, for example, use CFD as a tool to decide the optimal sensor location a more accurate model and more information regarding the boundary conditions is needed. Copyright (C) 2001 John Wiley & Sons, Ltd.