Class analytics

Students down one axis, gates across the other, mastery as colour. Columns follow the knowledge graph sequence rather than quiz order, so a cold column reads as a position in the course rather than as one bad assessment.

• A cold column is a teaching problem — the whole cohort missed that gate.
• A cold row is a student problem — remediation is already assigning them work.
• A cold block where a column meets a prerequisite is a sequencing problem: they never held the earlier gate.

Students flagged by a combination of mastery, trajectory, attendance and engagement rather than by a single failed assessment. The point is timing: a student identified in week four can still be helped; the same student identified by a final exam cannot.

The concept graph for the course with cohort performance mapped onto it, plus misconception clusters derived from wrong-answer patterns. Because MCQ distractors encode specific misconceptions, a cluster names what students actually believe rather than reporting that a question was difficult.

Class trajectory shows whether mastery is improving, flat or falling over time. Per-lesson analysis attributes movement to specific sessions — useful for spotting the lesson where a cohort quietly lost the thread.

From these signals LEAP proposes actions: refine a lesson, delay a gate, add remediation, shift the Bloom distribution, add a misconception check, slow the pacing. Each is accept, edit or reject — accepting applies it, editing lets you change it first.

Suggestions are proposals with a status, not automation. Nothing changes your course until you accept it.