The monitoring of Cr(III) and Cr(VI) in natural waters is essential for the assessment of chromium-based levels that deteriorate the water quality and adversely affect aquatic biota. The evaluation of the uncertainty of this monitoring is crucial for the determination of their fitness for purpose and for the objective interpretation of the quantitative information. There is a great need for a detailed bottom-up evaluation of the uncertainty of Cr(III) and Cr(VI) measurements in natural waters by ion-exchange high-performance liquid chromatography inductively coupled plasma mass spectrometry. Detailed models of the uncertainty of sample dilution, calibrator preparation and sample signal interpolation in a weighted calibration curve are proposed. The models proved adequate from the metrological compatibility between reference and estimated concentrations of spiked river waters. Measurement of Cr(III) and Cr(VI) above 0.98 µg L-1 and 0.5 µg L-1 with relative expanded uncertainties as low as 4.0%, and 2.3%, respectively, can be achieved. For not diluted samples, calibrators’ values and sample signal interpolation are responsible for between 14% and 82% or 18% and 85% of the uncertainty. The analysis of a certified reference material proved that the Cr(III)/Cr(VI) interconversion during analytical work is negligible. The developed measurement models were successfully applied to the analysis of river waters and groundwater collected in the Czech Republic. Only Cr(III) at the concentration levels around 1 µg L-1 was detected in the river samples, whereas a toxic Cr(VI) was determined in groundwater. The proposed validated methodology and user-friendly spreadsheet for the detailed evaluation of measurement uncertainty can be generally applied to other speciation approaches, where the separation techniques are hyphenated with either organic or elemental mass spectrometry.