Attempts (bars) and accuracy (line), per day. Accuracy is the share of recorded attempts answered correctly that day.
Every cluster with recorded attempts, best first. Bar length is accuracy; the number at the end is how many attempts it rests on.
Volume and accuracy side by side. A subproduct with high volume and low accuracy is either hard or badly written.
Worst first-attempt accuracy, not worst overall — the second try is contaminated by having just seen the answer.
How much of the bank anyone has ever touched. Content nobody opens costs the same to maintain as content everybody opens.
The level at which the bank is planned, so it is the level at which gaps are worth acting on.
Sortable. First is first-attempt accuracy; the gap between it and overall accuracy is how much the explanation is teaching.
The column that matters is top wrong pick: when one wrong option dominates, either it is a good distractor or the item is ambiguous. Items with fewer than 3 attempts are hidden — below that, accuracy is noise.
Accuracy by attempt number on the same item. If it does not climb, the explanation is not doing its job.
Which language the bank is actually consumed in — this is what decides where translation effort goes.
Hour of day by weekday, in UTC. Darker blue is heavier use.
Every recorded action, not just answers: opening a screen, flipping a card, skipping, revealing. leave carries the time on screen.
Every choice in Night Float asks Sure, Think so or Guessing. A good student is right more often when sure; "sure and wrong" is the dangerous corner.
Recognize, act, titrate, reassess, escalate. Under each bar: answers and the median simulated minutes from the call (or the previous step) to that choice.
Accuracy per organ system of the problem.
The only difficulty setting. Accuracy should fall as overlap rises.
Lowest accuracy first, with at least 3 answers each. "Harmful" counts the choices that hurt the patient.
Paste the code a student sent you — the one watermarked on every page. To put a name to a code, click the code badge at the bottom-right of any tool.
Click a code for the full report — the same screen the student sees. ★ marks a fictitious student (seeded for the launch; students never see this mark).
Read this before the charts. These counts say how much did not get recorded, so the numbers next door can be read with the right discount.