ERROR LOG + ANALYTICS-DRIVEN REMEDIATION

ALL YEARS | PAGE 37C - CAPTURE -> TAG -> REPAIR -> RE-TEST -> RECALIBRATE

An error log is useful only when it changes future behavior. The objective is to reduce repeat errors on new problems; a beautifully organized spreadsheet that does not alter the next study session is documentation, not remediation.

F. ERROR TAXONOMY - DIAGNOSE THE FAILURE

Error type	What happened	Best repair
Knowledge / retrieval	Fact, mechanism, association or formula was absent or unavailable without cues.	Targeted reteach -> closed-book retrieval -> new item after delay.
Reasoning / discriminator	Facts were known but the causal chain, localization or decisive clue was misused.	Compare top 2 choices; write the discriminating rule; practice look-alike cases.
Management sequence	Correct action chosen at the wrong time or after a missed stabilization/test step.	Write the algorithm as NOW -> NEXT -> DEFINITIVE -> DISPOSITION.
Question-reading / task	Lead-in, timeframe, age/setting, negation or key modifier was missed.	Underline task + time; summarize what is being asked before options.
Data / graph / calculation	Axes, units, direction, denominator or interpretation failed.	Reconstruct the relationship; solve 2-3 fresh representations.
Image / practical	Recognition depended on one memorized image, angle, slide or specimen.	Re-test with new views / modalities; verbalize defining features.
Timing / attention	Slow processing, rushed final items, careless changes or fixation.	Timed micro-sets; establish stop rules; track late-block decline.
Overthinking / low confidence	Moved away from the best-supported answer without new evidence or guessed correctly.	State the decisive evidence before changing an answer; review unstable corrects.

G. THE 48-HOUR REPAIR LOOP

Step	Execution
1. Capture	Record only misses/uncertain corrects with educational value. Save the tested rule, not the entire explanation.
2. Tag	Choose one primary error mechanism. If everything is tagged "knowledge," the taxonomy is not being used.
3. Repair	Use the smallest intervention that fixes the cause: one paragraph/video segment, one algorithm, one redraw or one worked example.
4. Retrieve	Close the source and reproduce the repaired rule from memory.
5. Re-test	Within ~24-72 hours, use a NEW vignette/image/problem. Recognition of the old answer does not count.
6. Escalate / archive	If the same error recurs, deepen the repair. If it transfers correctly twice, reduce review frequency.

H. WEEKLY DASHBOARD - WHAT TO TRACK

Metric	Why it matters	Trigger for action
Repeat-error rate	Best signal that remediation is failing.	Same mechanism misses twice -> targeted re-teach + fresh set.
Low-confidence correct rate	Detects unstable knowledge hidden by % correct.	High guessing -> retrieval / discriminator work.
Time per item + late-block drop	Separates knowledge from endurance/pacing.	Accuracy falls in final quartile -> timed stamina practice.
Domain trend	Shows decay and blind spots across systems/rotations.	Weak domain persists across multiple sets/assessments -> increase weighting.
Resource-to-output ratio	Prevents passive overconsumption.	Video/reading time rises while retrieval/questions stagnate -> prune resources.

H. WEEKLY DASHBOARD - WHAT TO TRACK — MONTHLY RECALIBRATION Use a simple STOP / START / CONTINUE: stop low-yield resource duplication; start one intervention for the largest repeat-error cluster; continue the

few habits producing durable retrieval and improving timed transfer. 2026/27 study-system expanded edition - verify school and official exam updates before testing