Student engagement in online learning remains a challenge for the design of effective coursework. Additionally, few analyses have focused on student-produced activities in the online mode or upon how such class activity affects student subgroups differently. We conducted a randomized design experiment with student video production at a large public university. Student background and behavior factors were measured in two online surveys, which were combined with course assessment data. Because of the small sample size, we observed few significant differences in learning outcomes across the experimental treatment and control sections, except with regard to a value-added measure. We suggest that student learning was likely most concentrated on concepts around which students produced the videos. And when students were divided by incoming language proficiency, non-native English speakers had higher perceived learning; but when grouped by incoming GPA, those with higher previous grades actually achieved higher test scores and pass rates.