The bibliometric measurement process transfers citations in scientific publications into indicators on scientific impact. However, the diverse measurement paths result in varying measurements and the identification of an optimal path is inhibited by the not perfectly understood relation between citations and the latent construct of scientific impact. We consequently propose to compute several measurement paths and analyse the resulting measurement variation caused by the diverse measurement processes. Scientific publications are understood as knowledge claims, while the therein listed citations initially illustrate the publication's (subjective) embedding. Together they constitute a partial mapping of the current knowledge space, which varies as every mesaurement process relies upon a particular mapping. E.g. the choice of a database with its particular coverage of scientific publications or the length of the citations window to count citations, result in changes to the impact values of the analysed entities. These impact values bear nowadays important consequences like funding opportunities or individual promotion, but currently neglect any inherent measurement variation.