The uncertainties of stable isotope results depend not only on the technical aspects of measurements, but also on how raw data are normalized to one of the international isotope scales. The inconsistency in the normalization methods used and in the selection of standards may lead to substantial differences in the results obtained. Therefore, unification of the data processing protocols employed is highly desirable. The best performing methods are two-point or multipoint normalization methods based on linear regression. Linear regression is most robust when based on standards that cover the entire range of δ values typically observed in nature, regardless of the δ values of the samples analysed. The uncertainty can be reduced by 50 % if measurements of two different standards are performed four times, or measurements of four standards are performed twice, with each batch of samples. Chemical matrix matching between standards and samples seems to be critical for δ (18)O of nitrate or δ (2)H of hair samples (thermal conversion/elemental analyser), for example; however, it is not necessarily always critical for all types of samples and techniques (e.g. not for most δ (15)N and δ (13)C elemental analyser analyses). To ensure that all published data can be recalculated, if δ values of standards or the isotope scales are to be updated, the details of the normalization technique and the δ values of the standards used should always be clearly reported.