Heatmaps and scores that show where attention lives

Measuring Attention and Comprehension, Not Just Views

A completed play tells you almost nothing. The video may have run in a background tab, or been skimmed at double speed with the key section skipped. Annoto's attention and comprehension analytics separate the two questions every instructor actually has: did students genuinely watch this, and did they understand it. Those are different problems, and they need different fixes.

How the two metrics are built

Attention draws on watching behavior. Because Annoto runs on top of Kaltura, Panopto, Wistia, YouTube, and Vimeo players, it observes how a session unfolds: continuous viewing versus skipping, replayed segments, and the points where learners leave. Comprehension draws on responses, including in-video quiz answers, poll results, and participation in time-anchored discussion. Both roll up into per-learner and per-video views delivered inside Canvas, Moodle, Blackboard, Brightspace, and Open edX through LTI 1.3.

Where the metrics fit in teaching

The pairing works as an early-warning system. High attention with low comprehension points to a content problem: the material needs a clearer explanation or a worked example. Low attention points to an engagement problem, often solved by shortening the video or adding an interaction. At the learner level, a student whose attention drops across several weeks can be contacted while there is still time to recover, instead of being discovered in the final grade distribution.

What the data shows in practice

Segment-level patterns are the most actionable part. Repeatedly rewatched minutes mark confusion, sharp drop-offs mark pacing failures, and comparing the same video across cohorts shows whether a redesign worked. Many teams use the metrics to test interventions, checking whether adding an in-video quiz or a reflection point changes how the next group watches. Everything can leave the platform through analytics export as CSV or through the API.

Attention and comprehension analytics summarize what the rest of Annoto's layer produces, from discussions and quizzes to polls and notes, and completion and gradebook sync carries the outcomes into your LMS. Institutions worldwide rely on these metrics. Definitions and setup guides are available at docs.annoto.net.

Reading a heatmap without overthinking it

A heatmap rewards a quick scan more than a deep audit. Three shapes cover most of what appears. A slow, even fade across the timeline is normal fatigue and usually means the video is simply long. A sharp cliff at one timestamp is structural: an unclear transition, a hard concept introduced without setup, or the moment the lecture stops matching the slides. A bright rewatch band is the most valuable shape of all, because learners are voting with their scrub bar for the passage they need explained again. Pick the single strongest signal per video and act on that; chasing every ripple turns insight into busywork.

A routine that fits inside course prep

Instructors who get real value from these metrics tend to run the same short loop. Before preparing the next session, open the heatmap for last week's videos, note the strongest rewatch band, and open that passage's time-anchored discussion to see what learners asked in their own words. That pairing — where they struggled plus how they describe the struggle — usually writes the first ten minutes of the next class by itself. Once a term, the same review at course level decides which videos get re-recorded and which just need a quiz inserted at the failure point.

What the numbers cannot tell you

Restraint matters. In a small cohort, one student's odd afternoon can bend a curve, so patterns count and single spikes rarely do. Attention describes behavior, not effort — a learner may absorb a topic from the textbook and skim the video legitimately. That is why comprehension scoring exists as the second check: act on the pair, not on either metric alone, and treat an individual learner's data as the start of a conversation rather than a verdict. Used that way, the analytics build trust instead of surveillance anxiety.