Abstract Two geochemical surveys of Pb in the environment were assessed for random errors originating from the sampling and analytical procedures. The levels of measurement error that were considered acceptable were set by comparison with the geochemical variation in the individual survey, rather than at a predetermined absolute threshold. Robust analysis of variance was used to separate the analytical and sampling errors from the “true” geochemical variation in the survey areas. Robust statistics were shown to be much more reliable than classical statistics for such environmental surveys where there are usually a small number of outlying values. A limit of 20% was set as the maximum acceptable proportion of the total variance that could be contributed by the measurement processes. A stream sediment survey for Pb showed a 4.3% contribution from measurement variance, whereas an urban soil survey gave a corresponding value of 43.3%. The distortion evident in the geochemical information for the soil traverse confirms the suitability of this approach to Sampling and Analytical Quality Control (SAX).