Remote maintenance will determine the available uptime of future fusion plants such as ITER. Experience at predecessor JET showed that a human-in-the-loop tele-operated approach is crucial, although this approach entails drawbacks such as the unavoidable extensive operator training and relatively long execution times. These drawbacks are common knowledge, but little quantitative research is available to guide improvements (such as improved training methods, or active operator support systems). The aim of this paper is to identify the key areas for further improvement of tele-operated maintenance. This is achieved by a detailed task analysis based on recent maintenance at JET, using task logbooks and video data as well as interviews with experienced master–slave operators. The resulting task analysis shows the (sub)tasks that were most time-consuming and shows a large variance in time performance within operators, but also substantial differences between qualified operators with different levels of experience. The operator interviews indicate that intuitive (virtual) visual feedback and artificial (guiding) forces are promising directions for improvement. The results found in this study will be used for future research and development activities focusing on haptic guiding strategies, with the aim to further design and optimize RH maintenance systems for ITER and beyond.