The aim of the Karlsruhe Tritium Neutrino experiment (KATRIN) is the direct (model-independent) measurement of the neutrino mass. For that purpose a windowless gaseous tritium source is used, with a tritium throughput of 40 g/day. In order to reach the design sensitivity of 0.2 eV/c² (90% C.L.) the key parameters of the tritium source, i.e. the gas inlet rate and the gas composition, have to be stabilized and monitored at the 0.1% level (1σ). Any small change of the tritium gas composition will manifest itself in non-negligible effects on the KATRIN measurements; therefore, Laser Raman spectroscopy (LARA) is the method of choice for the monitoring of the gas composition because it is a non-invasive and fast in-line measurement technique. In this paper, the requirements of KATRIN for statistical and systematical uncertainties of this method are discussed. An overview of the current performance of the LARA system with respect to precision will be given. In addition, two complementary approaches of intensity calibration are presented.