Collaborative Human-Robot workcells introduce robot-assisted operations in small-volume production or assembly processes, where conventional automation is noncompetitive. Unfortunately, the collaborative work of humans and robots sharing the same work area and/or working on the same assembly operation may pose unprecedented problems and failure risks. Failure Mode, Effects and Criticality Analysis (FMECA) is a popular tool to design reliable processes, which investigates the potential failure modes from the perspective of severity, occurrence and detection. The traditional FMECA approach requires the assessment of failure modes to be carried out collectively by a group of experts. Nevertheless, in the field of Human-Robot collaboration, experts are often unlikely to agree in their judgements, due to the almost inexistent historical records. Additionally, the traditional approach is not appropriate for decentralized production/assembly processes.The paper revisits the traditional approach and integrates it with the ZMII-technique – i.e., a recent aggregation technique developed by the authors – which overcomes some limitations, including but not limited to: (i) arbitrary categorization and questionable aggregation of the expert judgments, (ii) disregarding the variability in these judgments, and (iii) disregarding the result uncertainty. The description is supported by a real-life application example.