A weak quiz tests memory for 30 seconds. A strong quiz proves the employee can apply the rule on Monday morning. The difference is not difficulty — it is construction.
This guide breaks quiz-writing into five steps, draws on 31 validated item-writing rules, and shows two worked examples for SMB teams.
Start with what you measure
Before the first question, write one sentence: what does the employee do differently after the module than before. That sentence is the learning objective. Without it, the quiz measures vocabulary, not behaviour.
Bloom's revised taxonomy has six levels. In B2B SMB training, three drive value:
- Apply: the employee executes the right step in a scenario.
- Analyze: the employee picks out which rule fits the situation.
- Evaluate: the employee compares two options and selects the compliant one.
Leave Remember and Understand for an opening warm-up, not for impact measurement. "What does GDPR stand for" tests vocabulary, not behaviour.
Five rules for the stem
The stem is the question itself. Haladyna, Downing and Rodriguez codified 31 rules across 27 textbooks and 27 research studies. Five matter most for corporate use:
- One goal, one question. No "What is a breach AND who reports it?". Split it into two items.
- Write stems in positive form. Negative stems ("Which is NOT…") raise error rates even among learners who know the material.
- Put context in the stem, not the options. Reading the options should take less time than reading the stem.
- Avoid lifting words from the material verbatim. "Under Article 33 GDPR…" in the stem and "Article 33 GDPR" in the right answer is a grammatical cue.
- No absolutes. Words like "always", "never", "all", "none" in options signal the wrong answer before reading.
Distractors: where the quiz wins or loses
Write the correct answer first. Distractors — the wrong options — are the hard part. Three signs of a strong distractor:
- It looks plausible to a learner who hasn't prepared.
- It mirrors a real-world misconception, not a fabricated one.
- Its length is within ±20 % of the correct option.
Three signs of a weak distractor:
- It is funny or visibly wrong ("Call the president").
- It is shorter or longer than the rest.
- It does not match the stem grammatically (singular vs. plural, tense mismatch).
Rule of thumb: after the first 30 attempts, any distractor with 0 % selection becomes a giveaway. Replace it. Three real options beat four where one is decorative.
Pass threshold, remediation, length
Industry benchmark: 82 % of organisations require a passing score to mark a module complete. For compliance topics, 90 % is the target; for onboarding, 80 %; for soft skills, 70 %.
When does a low score mean the quiz is broken, not the learners? If average correctness on a single question falls below 60 % across the team, the question is unclear or the material did not cover it. Do not fail the learners; rewrite the question.
Length:
- 5-minute module: 3-5 questions.
- 10-minute module: 6-8 questions.
- 20-minute module: 10-12 questions, no more.
Long quizzes raise abandonment. In shorter microlearning modules, five questions is the sweet spot.
Worked example 1: 50-person SaaS, GDPR onboarding
A 50-person SaaS startup. Goal: after onboarding, every employee recognises a data breach and knows the escalation chain. Without that, the GDPR training program does not hold up under audit.
Module: 6-minute video + 5 questions.
Question 1 (Apply): "Your laptop with customer emails goes missing on a train. What is your first step in the next hour?"
- A) Notify IT and inform the DPO. ✓
- B) Call the customers and apologise.
- C) Wait until the Monday standup.
- D) Change passwords and continue working.
Question 2 (Evaluate): "A marketing colleague wants to email a newsletter to 5,000 subscribers who consented in 2019. Which statement is correct?"
- A) 2019 consent covers any future marketing purpose.
- B) Check whether that consent covered this specific purpose under Article 7. ✓
- C) After 5 years, no restrictions apply.
- D) Ask an external lawyer for written approval.
After three months: average first-try score 87 %. Question 4 dropped to 52 %. HR rewrote it — added one sentence of context. Next quarter: 81 %.
Worked example 2: 150-person distributor, EU health and safety
A 150-person distributor, 90 in the warehouse. EU member-state health and safety rules require risk-based training and a periodic refresher — the supervisory authority will ask for proof.
Module: 8 minutes + 7 questions. Separate scenarios for warehouse and office.
Warehouse question (Apply): "A forklift driver sees a wet floor in a transport aisle. What does she do first?"
- A) Drives around and continues.
- B) Calls cleaning and waits without marking the area.
- C) Marks the area and notifies the shift lead. ✓
- D) Ends her shift and goes home.
Office question (Analyze): "After 6 hours of uninterrupted screen work, the 20-20-20 rule says the employee should:"
- A) Shut down the computer.
- B) Every 20 minutes, look at an object 6 metres away for 20 seconds. ✓
- C) Work until end of shift.
- D) Change the chair.
EU law requires a training record. HR exports the LMS log — name, date, score — straight into the personnel file. Inspectors ask for exactly that.
Five common quiz-writing mistakes
Five patterns repeat in HR quizzes that miss the mark:
- "All of the above" as an option. A learner without knowledge picks it and lands the right answer roughly 60 % of the time.
- Stem at the end after two or three sentences of background. Reading takes longer than answering.
- The correct option is always the longest. A test-wise learner spots the pattern after three questions.
- Humour in a distractor. It breaks tension and reveals the right answer.
- No explanation on a wrong answer. The learner repeats the mistake because they don't know why it was wrong.
Each of these surfaces in a 30-minute pre-publish review.
Where this goes next
A good quiz takes longer to write than the module content itself. Five questions built on Haladyna's rules with verified distractors take HR more than one hour. An LMS like Mentor ships question templates, a reusable item bank, and the completion log inspectors expect, but it does not write the questions for you. The gap between "a quiz" and "a quiz that measures" is the 31 rules and two rounds of review before publish.