How to Write Better Exam Questions and Rubrics

A good assessment measures the learning you intended, fairly and consistently. That is harder than it looks — here are the principles that make it work.

Start with constructive alignment

Every assessment should trace back to a learning outcome. The principle of constructive alignment says your outcomes, your teaching activities, and your assessment must all point at the same target. If an outcome says students will ‘evaluate competing strategies’, a multiple-choice question asking them to recall a definition does not assess it. Before writing any item, list your outcomes and the cognitive level each demands — remember, understand, apply, analyse, evaluate, create — and write items at that level.

Write multiple-choice items that test thinking

Good MCQs can assess far more than recall, but most are written badly. The rules:

• Put the whole problem in the stem. The question should be answerable before the student reads the options.
• One unambiguously correct answer. If experts could argue, rewrite it.
• Plausible distractors. Wrong options should reflect real misconceptions, not filler. A distractor no one picks teaches you nothing.
• Avoid ‘all of the above’, absolute words (always, never), and grammatical giveaways that signal the key.
• Test application. Use a short scenario and ask students to apply a concept, not name it.

Design essay and short-answer prompts that are answerable

An essay question should make the task and the criteria obvious. Use precise verbs that match the outcome (‘compare’, ‘justify’, ‘critique’) rather than vague ones (‘discuss’, ‘write about’). Tell students the scope and weighting. A prompt that ten students interpret ten different ways will produce answers you cannot grade consistently — and complaints you cannot defend.

Build a rubric before you grade, not after

A rubric is a contract: it tells students what good work looks like and tells you how to grade fairly. An analytic rubric breaks the task into criteria (argument, evidence, structure, mechanics), each with described performance levels. Write the level descriptors in observable terms — ‘claims are supported by specific evidence from at least two sources’ beats ‘good use of evidence’. Share the rubric with students before the assessment; transparency improves work and pre-empts disputes.

Reduce bias and keep grading consistent

• Grade question-by-question, not student-by-student, so your standard for each item stays stable.
• Anonymise where possible to limit halo effects.
• Calibrate by grading a few scripts against the rubric first, and re-checking your earliest grades at the end — standards drift as you tire.
• Spot-check borderline cases against the rubric rather than against other students.

Check the assessment after the fact

After grading, a quick item analysis tells you whether the assessment worked. If almost everyone missed a question, it may be mis-keyed, badly worded, or untaught. If a question failed to separate strong from weak students, it may be too easy or ambiguous. Treat each exam as data about your teaching, not just about your students.

Final thought

Fair assessment is designed, not improvised: align to outcomes, write items at the right cognitive level, build the rubric before you grade, and grade in a way that controls bias. Assessment Studio accelerates the drafting so you can spend your time on the judgement that makes an assessment fair.