To justify animal experiments, all methods require State-of-the-Art techniques. Statistics belong to often ‘unloved’ methods but reduce subjects by biometrical predictions, statistical analyses, and, surprisingly, by correct identification, i.e. the experimental unit (EU). Per definition, the EU is the number of subjects that can be treated individually different within the experiment. Consequently, correct implementations of EUs increase animal numbers when ‘treatment’ applies grouped housed animals, because the cage, independent of number of animals, is the EU. Thus, either ignore sociability and single-house or testing one single cage-representative, thereby producing several ‘surplus’ animals or testing all animals of a cage (but using only the mean of the cage) would be a solution. To analyze these 4 optional conditions (single-housed, ‘cage-representative’, ‘cage mean’ or all individual within cage), we analyzed male and female C57BL/6NRj and female SWISS mice concerning animal welfare behavior and physiology, thereby regarding activity, anxiety, nociception, exploration and corticosterone. Hierarchy was weekly assessed to correlate behavior with rank. Several significant, but not striking effects of condition could be observed. Coefficient of Variation was not influenced by condition but dependent on the parameter assessed. Hierarchical ranks were rather stable within the groups and did not correlate with the behavioral phenotype. Concluding, application of the EU as either ‘mean’ or ‘representative’ of the cage does not differ from individual analyses of mice within a cage. Therefore, animals’ social needs can be honored over statistical needs resulting in higher animal welfare and more respect for the individual life of a mouse.