AbstractA common approach to determining distances to stars without astrometric information is to compare stellar evolution models with parameters obtained from spectroscopic techniques. This method is routinely applied in the context of large-scale stellar surveys out to distances of several kpc. However, systematic errors may arise because of inaccurate spectroscopic parameters. We explore the effects of non-local thermodynamic equilibrium (NLTE) on the determination of surface gravities and metallicities for a large sample of metal-poor stars within approximately 10 kpc of the Sun. Using the improved T_eff scale, we then show that stellar parameters estimated based on the widely used method of 1D LTE excitation-ionization balance of Fe results in distances which are systematically in error. For metal-poor giants, [Fe/H] ~ −2 dex, the distances can be overestimated by up to 70%. We compare the results with those from the Radial Velocity Experiment Survey catalogue (rave) for the stars in common, and find similar offsets.