The question bank is the most underrated asset in any assessment programme. Institutions invest heavily in exam delivery technology but often neglect the quality of the questions themselves. Here are 10 practices that distinguish high-quality banks from the rest.
1. Tag every question with difficulty, topic, and learning objective
Without structured metadata, you cannot build balanced exams programmatically. Every question should have at least: difficulty level (Easy/Medium/Hard), topic/sub-topic, and the learning objective it tests.
2. Run item analysis after every exam
Post-exam analysis reveals two critical statistics: Difficulty Index (proportion of students who got it right) and Discrimination Index (whether the question separates high performers from low performers). Any question with a discrimination index below 0.2 should be reviewed or retired.
3. Avoid single-difficulty exams
A well-designed exam typically follows a 30/50/20 split: 30% easy (confidence-building), 50% medium (core assessment), 20% hard (differentiation). Building this intentionally requires a tagged question bank.
4. Use question variants to prevent leakage
For high-stakes exams, create 3–5 variants of each question that test the same concept but with different numbers, scenarios, or phrasing. Randomise which variant each candidate sees.
5. Retire questions after heavy use
A question seen by more than 2,000 candidates in a coaching context is likely to be circulating in WhatsApp groups. Track exposure and retire questions proactively before they become memory items.