A new dataset of uniform and steady sheet-flow experiments is presented in this paper. An acoustic concentration and velocity profiler (ACVP) is used to measure time-resolved profiles of collocated 2C velocity ( $u,w$ ) and sediment concentration and to measure the time evolution of the bed interface position. Ensemble averaging over 11 similar experiment realisations is done to evaluate the mean profiles of streamwise velocity, concentration, sediment flux and Reynolds shear stress. The repeatability, stationarity and uniformity of the flow are carefully checked for a Shields number ${\it\theta}\approx 0.5$ and a suspension number of $S=1.1$ . The mean profile analysis allows to separate the flow into two distinct layers: a suspension layer dominated by turbulence and a bed layer dominated by granular interactions. The bed layer can be further subdivided into a frictional layer capped by a collisional layer. In the suspension layer, the mixing length profile is linear with a strongly reduced von Karman parameter equal to 0.225. The Schmidt number is found to be constant in this region with a mean value of ${\it\sigma}_{s}=0.44$ . The present results are then interpreted in terms of existing modelling approaches and the underlying assumptions are discussed. In particular, the well-known Rouse profile is shown to predict the concentration profile adequately in the suspension layer provided that all the required parameters can be evaluated separately. However, the strong intermittency of the flow observed in the bed layer under the impact of turbulent large-scale coherent flow structures suggests the limitations of averaged steady two-phase flow models.