The robustness of the popular 'Tanaka' LC column characterization protocol has been evaluated by the statistical tools of reduced factorial design, multiple linear regression and principal component analysis. These have shown that in order to obtain reliable and reproducible results, it is especially important to control the methanol content, the temperature and, in the case of the total ion-exchange capacity test (alphaB/P pH 7.6), the pH of the mobile phase. In particular, the hydrophobicity tests (kPB and alphaCH2) are sensitive to small changes in methanol content. Provided that the operating parameters for the Tanaka column characterization protocol are controlled within the following experimental limits, i.e. methanol content +/-0.5% v/v, temperature +/-3degreesC, pH +/-0.10 and buffer concentration +/-2.0 mM, it is feasible to distinguish between RP materials that possess selectivity differences larger than their batch-to-batch reproducibility. These experimental requirements can be easily met by current LC instrumentation. Hence, the validity of the Tanaka testing protocol for characterizing columns has been verified.