Practical aspects concerning the use of 3D Navier-Stokes solvers as prediction tools for micro-siting of wind energy installations are considered. Micro-siting is an important issue for a successful application of wind energy in sites of complex terrain. There is a constantly increasing interest in using mean wind flow predictions based on Reynolds averaged Navier-Stokes (RANS) solvers in order to minimize the number of required field measurements. In this connection, certain numerical aspects, such as the extent of the numerical flow domain, the choice of the appropriate inflow boundary conditions, and the grid resolution, can decisively affect the quality of the predictions. In the present paper these aspects are analyzed with reference to the Askervein hill data base of full scale measurements. The objective of the work is to provide guidelines with respect to the definition of appropriate boundary conditions and the construction of an adequate and effective computational grid when a RANS solver is implemented. In particular it is concluded that (a) the ground roughness affects the predictions significantly, (b) the computational domain should have an extent permitting the full development of the flow before entering the region of interest, and (c) the quality of the predictions at the local altitude maxima depends on the grid density in the main flow direction.