YOUR QUESTION BANK
See how your exam questions are doing.
A report card for every question you've used, built from your own exam results.
Recent exams
Class averageWorth a second look
Signals, not verdictsReady for your own exam?
Add your question list and your students' scores. You'll get a report card for every question, and you can save the results as a file to reopen next term.
| Question | Topic | Times used | Last used | Average score | Separation |
|---|
Click a question for every exam it appeared on. Average score is the share of marks the class earned; separation shows whether stronger students did better on it.
Your question list
Add it once, then again whenever you write or rewrite questions. One row per question, with the marks it's worth and its topic.
No question list added yet.
What should the file look like?
One row per question. Keep the code the same when you rewrite a question, and give the rewrite a new version number.
| item_id | version | title | max_points | outcome_id |
|---|---|---|---|---|
| FIN-01 | v1 | Calculate a break-even point | 1 | Finance |
| MKT-01 | v2 | Choose a target market (short case) | 2 | Marketing |
- item_id: a short code you choose, like FIN-01. It stays the same across rewrites.
- version: v1 for the original, then v2, v3 for each rewrite.
- title: a short description, not the full question.
- max_points: the marks the question is worth.
- outcome_id: its topic or learning outcome.
- Optional: concept (a short name) and source_reference (for example, a textbook chapter).
An exam's scores
Tell us which exam this was, then add each student's marks on each question. Names and student numbers in the file are ignored.
Topic balance on your most recent exam
Share of marks, not masteryQuestions you could use next
With each one's track recordWhat the pilot does
It keeps a record of every question you've used: which exams it appeared on, which version ran, and how the class did. It works out the numbers itself from your score files, the same way every time, whichever platform ran the exam.
How to read the numbers
- Average score: the share of the available marks the class earned. A 2-mark question that averages 1.5 marks scores 75%.
- Separation: whether students who did well on the rest of the exam also did well on this question. Above zero, stronger students did better. Below zero, they did worse, which often points to a problem with the answer key.
- Worth a second look: questions where stronger students did worse, or where almost everyone, or almost no one, got it right. These are signals. The decision is always yours.
- Original and Rewrite: when you rewrite a question it becomes a new version. Each version keeps its own results, and "last used" counts every version.
What it never stores
Student names, student numbers and individual scores. Files are read in your browser, used once and discarded. Saved projects hold question details and summary results only. Small-class results can still be sensitive, so store project files with care, and use the sample class until your institution has approved real exam data.
What it doesn't do yet
Import directly from Moodle or Canvas, store full question text and answer keys, build exams for you, or share a bank with colleagues. Those are on the roadmap.
For the statistically minded
- Average score (facility)
100 × mean(points) ÷ max points, among students who were scored on the question.- Separation (corrected item-rest correlation)
- Pearson correlation between a question's points and each student's total on the rest of the exam. Reported only with at least 10 complete attempts on a common form, and withheld when there is no variation.
- Top-versus-bottom gap (upper–lower 27% index)
- Mean proportion of marks for the top 27% on the rest of the exam minus the bottom 27%. Withheld when tied rest-of-exam scores straddle a group cutoff, rather than splitting tied students arbitrarily.
- Class average, median and spread
- Percent of total marks; spread is the sample standard deviation. Reported only when every student wrote the same questions.
- Flags
- Negative separation, facility above 95% or below 20%, and fewer than 30 respondents. Flags never remove a question.
- What is never claimed
- Results are not pooled across exams, and similar averages are not treated as evidence that two exams are equivalent.
Calculation method: . Design decisions are documented as architecture decision records in docs/adr/ in the source repository.