Inadequate consideration of course workload in undergraduate students' course selections has contributed to adverse academic outcomes. At the same time, credit hours, the default institutional metric to convey time-based course workload to students, has been shown to capture students' experienced workload insufficiently. Recent research documents the accuracy of analytics learned from enrolment and learning management system data that promise to convey course facets other than time to students. To investigate the feasibility of scalable interventions of such analytics on course selection, we conducted a preregistered and randomized forced-choice experiment of two courses satisfying equal requirements across 61 undergraduate students at a large public university. The control condition mimicked a university's standard course catalogue, while the intervention condition augmented this catalogue with analytics on time load, mental effort, and psychological stress alongside institutional credit hours. We find that course load analytics (CLA) significantly influenced students' course selection decisions, leading them to more deliberately optimize their expected workload relative to credit hours. Specifically, we document a strategy regarding students being more likely to take a higher credit hour course that minimizes the expected workload associated with each credit hour. Among all three dimensions of workload, students' course selection decisions were most guided by predicted mental effort, followed by psychological stress, highlighting the limitations of a time-based credit hour metric. The intervention effect of CLA was constant across first-year students and comparatively senior students. We further discuss the implications of our findings for CLA interventions at scale and the institutional stakeholders that could be involved.