Following a brief historical background, the concepts and the present state of sputter–depth profiling for thin–film analysis are outlined. There are two main branches: either the removed matter (as in mass– or optical–spectroscopy–based secondary–ion mass spectrometry or glow–discharge optical emission spectroscopy), or the remaining surface (as in Auger electron spectroscopy and X–ray photoelectron spectroscopy) is characterized. These complementary methods show the same result if there is no preferential sputtering of a component. The common root of both is the fundamental ion–solid interaction. Understanding of how the latter influences the depth resolution has led to important improvements in experimental profiling conditions such as sample rotation and the use of low–energy ions at glancing incidence. Modern surface–analysis instruments can provide high–resolution depth profiles on the nanometre scale. Mathematical models of different sophistication were developed to allow deconvolution of the measured profile or quantification by reconstruction of the in–depth distribution of composition. For the latter purpose, the usefulness of the so–called mixing–roughness–information (MRI) depth model is outlined on several thin–film structures (e.g. AlAs/GaAs and Si/Ge), including its extension to quantification of sputter–depth profiles in layer structures with preferential sputtering of one component (Ta/Si). Using the MRI model, diffusion coefficients at interfaces as low as 10−22 m2 s−1 can be determined. Fundamental limitations of sputter–depth profiling are mainly traced back to the stochastic nature of primary–particle energy transfer to the sputtered particle, promoting atomic mixing and the development of surface roughness. Owing to more sophisticated experimental methods, such as low–energy cluster ion bombardment, glancing ion incidence or ‘backside’ sputtering, these ultimate limitations can be reduced to the atomic monolayer scale.