Drop-off rate is the proportion of learners who stop watching a video at a given timestamp, usually expressed as a retention curve that starts at 100 % at the opening second and declines over the video's duration. In a learning context the curve has a characteristic shape: a steep initial drop in the first few seconds (learners who bounced before committing), a gradual decline through the body, and a second steeper drop wherever the content becomes confusing, slow, or technically broken. The raw data feeding this curve comes from playback events — played and paused timestamps or the "progress" field in xAPI Video Profile statements sent to an LRS (Learning Record Store) — aggregated across the cohort. The instructional value lies in interpreting spikes rather than the overall slope: a sudden cliff at minute three, where the average drop in a comparable video would be gradual, is a precise flag for an instructional problem at that point. A common misreading is treating any drop-off as failure; for long videos a gradual linear decline often means learners watched what they needed and left intentionally. Contextual data such as post-video quiz performance is necessary to distinguish purposeful exit from disengagement. Drop-off rate is most useful when compared across cohorts or time periods, revealing whether a content revision lowered the cliff or merely moved it to a different timestamp.
