This study investigates EFL students’ subtitling workflows for self-produced mini documentaries, drawing on Romero-Fresco’s (2013) accessible filmmaking framework and Tardel’s (2023) subtitling model. Participants are students in an elective English course at a Japanese national university where they conduct an interview in Japanese, edit the video and create reverse subtitles in English; authentic tasks that involve multimodal literacy skills. The research employs an in-depth qualitative case study design with thick description to ensure reliability (Riege, 2003), supplemented by quantitative measures. The pilot (n=1) failed when think-aloud protocols (TAPs), the original main instrument, proved unsuitable due to scheduling constraints, TAP limitations for complex multimodal tasks, and language barriers, yielding insufficient workflow data. For the main study (n=2 focal participants), we pivoted to triangulation (Yin, 2018): home-based extended screen recordings for ecological validity and depth, in addition to field notes, logbooks, learning diaries, and interviews. These are bolstered by materials from consenting classmates (n=20) to strengthen the findings. Preliminary analysis suggests this combination overcomes pilot failures, enabling richer data on subtitling processes and demonstrating how methodological setbacks can strengthen research design.