How to read test results like an exam board

1. Difficulty: how many people got it right

Difficulty is simply the percentage of people who answered correctly. Between about 30% and 85% is usually healthy. Above 95%, the question tells you almost nothing because everyone gets it. Below 20%, check the answer key first: the most common cause of a “very hard” question is a mistake in the key.

2. Discrimination: does it separate strong from weak?

Discrimination compares how people did on one question with how they did on the rest of the test. A good question is answered correctly more often by people who did well overall. Values above 0.30 are good; 0.15–0.30 is acceptable; below 0.15 the question isn’t doing much; and a negative value means strong candidates are getting it wrong. That almost always means confusing wording or a wrong answer key.

3. Distractors: which wrong answers people chose

For multiple-choice questions, look at how often each wrong option was picked. A wrong option nobody chooses should be replaced with a more realistic mistake. A wrong option that strong candidates choose more than weak ones may actually be correct, or the question may be ambiguous.

What to do with a problem question

• Too easy: make the wrong options more realistic, or replace it.
• Too hard: check the key and the wording, then decide whether the skill really belongs in the test.
• Low or negative discrimination: rewrite the stem so there is exactly one defensible answer.

Wait for at least 20–30 people before acting on these numbers; with only a handful, results bounce around. For overall test quality, read test reliability explained.