Banning or reduction of the use of animals for laboratory experiments is a frequently-discussed societal and scientific issue. Moreover, the usefulness of animals needs to be considered in any decision process on the permission of specific animal studies. This complex issue is often simplified and generalized in the media around the question, “Are animals useful as a model?” To render an often emotional discussion about animal experimentation more rational, it is important to define “usefulness” in a structured and transparent way. To achieve such a goal, many sub-questions need to be asked, and the following aspects require clarification: (i) consistency of animal-derived data (robustness of the model system); (ii) scientific domain investigated (e.g., toxicology vs disease modelling vs therapy); (iii) measurement unit for “benefit” (inte-grating positive and negative aspects); (iv) benchmarking to alternatives; (v) definition of success criteria (how good is good enough); (vi) the procedure to assess benefit and necessity. This series of articles discusses the overall benchmarking process by specifying the six issues. The goal is to provide guidance on what needs to be clarified in scientific and political discussions. This framework should help in the future to structure available information, to identify and fill information gaps, and to arrive at rational decisions in various sub-fields of animal use. In part I of the series, we focus on the robustness of animal models. This describes the capacity of models to produce the same output/response when faced with the “same” input. Follow-up articles will cover the remaining usefulness aspects.