<p>Sustainable manufacturing processes and the circular economy require addressing issues such as increasing generation of waste and depletion of natural resource. Therefore, a part of end-of-life (EOL) products require disassembled and reused partially or completely. However, manual disassembly tasks are still required because EOL products have different conditions and standards under which they are collected. Hence, the motion of the disassembly task should be analyzed to improve its efficiency. In this study, the motion data of the nut-loosening task, which is a major disassembly task, was obtained via video camera and motion capture. A new approach that integrates motion study and data science is conducted for motion data on disassembly tasks. Furthermore, their movements were analyzed using unsupervised learning methods. Finally, motion study and data science were compared and discussed for nut-loosening tasks. The answers to the response questions obtained from this study indicate that spontaneous learning through simple repetition might lead to convergence to bad posture. Therefore, workers should receive instructions from their supervisors to correct their movement habits and reduce their burdens. The goal of this research is to support workers’ skill acquisition through a case study on manual disassembly tasks. Furthermore, the proposed approach is expected to be applicable to other types of manual disassembly tasks beyond nut-loosening.</p>