Writing a quiz: 4 rules to make it really teach

You finish a module and quickly add ten questions: "What is VAT?", "List the five stages of the process." Learners click, score 8 out of 10, and forget everything the following week. The quiz was there for show. It can do much better, without more work.

Why a quiz helps people remember#

In 2006, two American researchers, Henry Roediger and Jeffrey Karpicke, compared two ways of studying a text: re-reading it, or taking free-recall tests on it. Straight afterwards, re-reading gave better results. But when participants were tested two days or a week later, those who had tested themselves remembered markedly more. The re-readers, meanwhile, felt more sure of their memory.

This is retrieval practice, or the "testing effect". Fetching information from your head fixes it better than seeing it again in front of your eyes. This study involves students and texts in a laboratory, so it does not promise the same gain for every course. It gives a solid direction, not a promise.

Rule 1: one question, one idea, one situation#

A good question fits in one sentence and tests only one thing. It places the person in a case, rather than asking for a definition to be regurgitated.

The wrong answers on offer count as much as the right one. Write them using the real mistakes your learners make, the ones you have heard in training. An absurd answer teaches nobody anything.

Rule 2: give feedback that explains#

Writing "Wrong answer" is a missed opportunity. Useful feedback says three things: why the chosen answer does not work, why the right one does, and where to review the point in the course.

Another point: let the learner retake the quiz. A second attempt, after reading the feedback, gives the learner time to correct their mistake.

Rule 3: come back often, in small touches#

A 15-question quiz at the end of a month of lessons is an exam. Four questions at the end of each module, then two or three questions that revisit an earlier module, is training. The researchers we cite note, incidentally, that more frequent classroom testing could improve retention after the course ends.

A simple outline: 3 to 5 questions per module, a recall quiz a week later, and a short summary at the end of the path. Also note the questions that the whole group gets wrong: those are where the course needs improving, not the learner.

Rule 4: get help from AI, but read every question#

Writing questions takes time. Kognisia's AI-assisted generation proposes a first set of questions, which you then edit. Saving an hour on writing is useful, on one condition: human review is mandatory, question by question.

An AI can write an ambiguous question, a wrong answer presented as right, or a question that is too easy. Check three things before publishing:

• the right answer is indeed correct, and it is the only one;
• the wrong answers look like real mistakes;
• the question is about your content, not about general knowledge.

If you want to see what comes after a quiz, that is, putting people in a situation, read our article on assignments and case studies. Progress tracking then shows you the questions where the group stumbles.

Frequently asked questions#

In summary#

A quiz teaches when it asks people to retrieve an answer, explains the mistake and comes back regularly. Write situation questions, review the ones the AI proposes, and keep them short. Next step: this week, rewrite three questions from your last course as situation questions.

Sources#

• Roediger and Karpicke, "Test-enhanced learning: taking memory tests improves long-term retention", Psychological Science, 2006, vol. 17, no. 3, pp. 249 to 255 (WashU record), accessed 13 October 2026.
• Association for Psychological Science, Observer: "Test-enhanced Learning", 2006, accessed 13 October 2026.

The Kognisia team
Kognisia designs a SaaS learning platform, with a training catalogue and white label as options.