Most Direct Numerical Simulation (DNS) of turbulent flow of generalized Newtonian (GN) fluids presented to date have shown significant discrepancy between experimental measurement and simulation. In addition to DNS, empirical correlations using different rheology models fitted to the same shear rheogram have also shown to give significantly different results. Important to note is that for turbulent flow predictions it is a common practice to use a shear rheograth which is measured at shear rates well below the values encountered in turbulent flows. This paper highlights the importance of obtaining high shear rate rheology in reducing these discrepancies. Further, it is shown that if high shear rate rheology is used in rheology characterisation, the choice of rheology model has little influence on the results. An important aside is that accurate prediction of laminar flow gives absolutely no confidence that a rheology model is acceptable in modelling the turbulent flow of the same fluid. From an analysis of instantaneous shear rates in the predicted turbulent flow field, the probability distribution of the non-dimensionlised shear rates in the near-wall region appears to collapse onto a universal curve. Based on this, we propose that the maximum shear rate required in rheology characterisation should be at least twice the shear rate corresponding to the mean wall shear stress. (C) 2016 Elsevier B.V. All rights reserved.