Quantification of surface nano-structures by angle-resolved XPS (ARXPS) is straightforward and works quite reliably for perfectly flat surfaces of amorphous materials. For rough surfaces, the interpretation of ARXPS is, however, complicated because the angular variation of the XPS peak intensity depends on the surface roughness. Even for ideally flat substrates, ARXPS analysis of laterally inhomogeneous surface structures grown on the surface is quite complex. The reason is that neighboring nano-clusters shadow the XPS peak intensity. The effect depends on cluster shape as well as the distribution of clusters on the surface. In addition the effects depend on the flatness of the underlying substrate. The interpretation of ARXPS then becomes quite complex. In the present paper, we have studied this problem by analyzing ZnO nano-clusters grown on substrates of SiO 2 and Al 2 O 3 . Thus we compared the results of quantification by the four techniques: ARXPS, XPS-peak shape analysis, Rutherford backscattering spectroscopy and x-ray fluorescence spectrometry. While the latter three techniques gave consistent results, the results of the ARXPS analysis were way off. This deviation is discussed in terms of the above-mentioned shadowing effect of neighboring clusters as well as roughness of the underlying substrates. Different normalization methods in the ARXPS analysis procedure are compared and it is found that some of the observed problems for the substrate peaks (but not for the peaks from the overlayer film) can be reduced by applying reference samples with similar roughness for normalization of the data. In conclusion, the ARXPS technique is very much dependent on surface roughness as well as on the morphology of the thin films. Thus for reliable quantification with ARXPS it is necessary to have independent knowledge on surface roughness as well as the distribution of islands of the thin films.