Micellar electrokinetic capillary chromatography and reverse-phase liquid chromatography methods were developed in order to perform robustness testing to determine the caffeine content in beverages. Both methods were fully validated and two different robustness approaches were applied. One-variable-at-a-time (OVAT) approach at eleven levels (0; +/-1; +/-2; +/-3; +/-4; +/-5 units) was carried out and compared with multi-variable-at-a-time (MVAT) approach at three levels (+/-1 unit per investigated parameter). Four analysts in two laboratories on two capillary electrophoresis and two RPLC equipments have tested the samples. Robustness was statistically analyzed using peak area, migration or retention time, symmetry, and resolution of caffeine and sulfacetamide as internal standard, and presented as R.S.D. values. The RPLC method was found to be more sensitive than the MEKC method. Both methods showed acceptable robustness level for OVAT approach, whilst MEKC method was more robust when the determination of real samples coming from different beverages was based on the MVAT approach.