This paper provides guidance on the process of establishing input data to safety and reliability engineering analyses when no or little field data exist, and expert judgment is required. Some recommendations are directly related to a discussion of basic requirements for scientific work. Further, two case studies are discussed in order to highlight some actual problem areas that are experienced when using expert judgment, and some recommendations for handling these problems are given. The first case describes how expert judgment was used to analyse the safe operation of an umbilical on a semisubmersible drilling rig, and the second case is related to establishing generic failure rates/probabilities for components of offshore safety systems.