The impact of globalization and technologization has continued to reverberate in the translation industry, causing seismic changes in the translation process, output, and client expectations. Consequently, several approaches and models have been adopted to align training with these changes. Prominent among these is project-based learning, where most research has focused on collaborative practices that mirror authentic translation scenarios in the industry. However, despite the high volume of research on collaborative translation within the training environment, two significant lacunae remain perceptible--there is limited research comparing collaborative teams and individual trainees, and the use of video ethnography as a research methodology is scarce. This study intends to bridge this gap by comparing the performance of collaborative teams and individual trainees in accomplishing several computer-assisted translation (CAT) tasks. The study quantitively analyzes data from video recordings of four collaborative teams and 25 individual trainees performing three computer-assisted task blocks, eight subtasks, and three sub-subtasks. The study investigates compliance with instructions, task completion time, translation output quality, terminology research, and workspace management. The findings reveal that, to different degrees, teams and individual trainees failed to comply with instructions (even though teams spent more time working on the tasks), and teams had higher assessment scores and also enjoyed more downtime. In contrast, individual trainees had less downtime and lower average assessment scores though they consulted more terminological resources than teams. Furthermore, correlations were established between the final assessment score, task completion time, and compliance with instructions. The findings informed our practical and pedagogical reflections on CAT course design, particularly with regard to the training outcomes for collaborative teams and individual trainees.