How to make an online quiz: scoring, feedback and examples

To make an online quiz, decide first whether it tests knowledge (answers are right or wrong) or matches people to a result (a personality type or a product), then write short questions with believable wrong answers, set points and a pass mark, choose when people see the correct answers, and take the quiz yourself on a phone before you share the link.

For a classroom or training check, a sound default is ten single-choice questions with three or four plausible options each, a pass mark of 70 or 80 percent, and correct answers shown after submission.

Step 1: Choose the kind of online quiz

Choose a knowledge quiz when answers can be marked right or wrong, and a personality or product-recommendation quiz when every answer is acceptable and the point is to match the person to a result. They are built differently, so do not mix them in one quiz.

For a personality or product quiz, write the results first. Decide on three to five results (four coffee styles, three membership levels), write a short description and a next step for each, then write questions whose options each point to one result. Show the result that collects the most answers, and decide in advance how to break a tie. If you want an email address, ask for it after the result is on screen; an email field in front of the result reads as a toll gate.

Step 2: Write questions that test one thing

A good quiz question tests one fact or skill, can be answered without looking at the options, and is hard only because of the subject, not because of the wording.

1. Write a complete question. A stem like "Phishing:" followed by four statements makes people work out what you are asking first.
2. Test one idea. If someone could know half the answer, split the question.
3. Test use when the job requires use. "A customer emails asking you to change their bank details. What do you do first?" tells you more than "What is social engineering?"
4. Avoid "All of the above". Someone who spots two correct options can choose it without knowing the third. If several answers are correct, use a select-all question.

The survey question types guide covers each type in more depth; the scoring advice here is specific to quizzes.

Step 3: Write distractors people could believe

A distractor is a wrong answer option, and a good one is a mistake that someone who half-knows the topic would actually make. When the wrong options are absurd, the quiz measures common sense instead of knowledge.

Take distractors from the misconception you hear most often, from the result of a typical error (forgetting to convert minutes to hours, applying a discount after tax), or from the right procedure for a different situation. Keep every option the same length and grammatical form as the correct answer, because test-wise people pick the longest, most qualified option when they do not know. Three plausible options beat four that include a filler: Michael Rodriguez's 2005 meta-analysis of 80 years of research, published in Educational Measurement: Issues and Practice, concluded that three options are usually enough.

Step 4: Set points and a pass mark

Give every question one point unless some matter more than others, and set the pass mark from what a pass must guarantee, after checking how much guessing alone could score.

Weight a question only when it covers something a person must not get wrong, or put those critical rules in a separate short quiz with a higher pass mark, so a strong score elsewhere cannot hide a dangerous gap. For select-all questions, one common rule gives a point for each correct choice, takes one away for each wrong choice and never goes below zero.

A worked example: ten single-choice questions at one point each plus one matching question worth five points gives 15 points, so an 80 percent pass mark means 12. Eight answers and four pairs right scores 12 and passes; nine answers and two pairs scores 11 and does not.

Guessing sets the floor. On a ten-question quiz, someone who guesses every answer has these chances of reaching 70 percent:

A true-or-false quiz with a 70 percent pass mark lets roughly one pure guesser in six through. With three or four plausible options, guessing alone almost never passes.

These are starting points, not research findings. After the first group, check that the people who passed are the ones you would trust with the topic, and adjust if they are not.

Step 5: Decide when people see the correct answers

Show correct answers immediately in practice quizzes, after submission in graded quizzes taken once, and only after the last attempt when retakes are allowed. Otherwise the second attempt measures memory of the answer key.

The explanation matters more than the key. Write one or two sentences on why the right answer is right and why the most popular wrong answer is wrong, and make the fail message say what to do next:

Correct answer: report it with the Report phishing button. Reporting lets the security team block the sender for everyone. Opening the link "to check" is the most common mistake: a fake login page can capture your password before anything looks wrong.

Fail message: You scored 10 of 15 points, and the pass mark is 12. Review the sections on reporting and passwords, then try again tomorrow.

Step 6: Set time limits and attempts

Add a time limit only when speed is part of what you are testing or when you need to stop people looking answers up. For most training quizzes, leave it off.

When you use one, allow about a minute per single-choice question and two or three minutes for calculations and scenarios, then add a quarter again for people reading in a second language or on a small screen. Show the timer, warn a few minutes before the end, and submit automatically so nobody loses the answers they gave. If your course has accessibility obligations, plan how people who need more time will get it; the W3C's Web Content Accessibility Guidelines cover time limits in the Timing Adjustable success criterion.

For graded quizzes, allow two or three attempts with a day between them and keep the best score. For certification, keep the limit low and do not show the key between attempts.

Step 7: Reduce guessing and answer sharing

You cannot stop a determined cheater in an unsupervised online quiz, but you can make guessing unprofitable and shared answers much less useful.

1. Use three or four plausible options instead of true or false.
2. Shuffle questions and options for each person.
3. Write two or three versions of each question and give each person a random subset from a question pool, so neighbors see different quizzes.
4. Use scenario questions whose answers cannot be found with a quick search.
5. Send each person their own access code instead of posting a public link.
6. Treat proctoring signals such as tab switches and pasted text as a reason to talk to someone, not as proof. People switch tabs to answer a message, too.

Step 8: Test the quiz and review each question

Take the quiz yourself twice before anyone else sees it, once answering everything correctly and once answering everything wrong. Check the score, the pass and fail messages, the timer and the retake rule, then delete those test attempts.

After the first group, review each question's results (often called item analysis). A question almost everyone gets right tells you little. One that fewer people get right than guessing would predict needs its answer key and wording checked. Replace any distractor nobody picks, and rewrite any question where your strongest scorers prefer a wrong option.

To learn how the training itself landed, pair the quiz with a short training effectiveness survey, and keep opinions separate from marks.

Common mistakes

• The correct option is always the longest one. Match the length and detail of every option.
• "All of the above" is the answer. Rewrite the question as a select-all question.
• A true-or-false quiz with a low pass mark. Add options or raise the mark.
• Correct answers shown immediately in a graded quiz with retakes. Show them after the last attempt.
• No review after the first group. Read the per-question results and fix the weakest questions.

Doing this in Claform

A Claform quiz is a form with its own quiz settings. You mark correct answers and points on choice, number, text, rating, ranking and matrix questions, and the quiz is scored on submit, with partial credit on matrix questions and a grading queue for open answers. The settings cover a pass mark with pass and fail messages, when correct answers and explanations appear, a timer with a warning and auto-submit, retakes with a wait between attempts, shuffled questions and options, question pools, personal access codes and proctoring signals such as tab switches and blocked pasting. Results go to a gradebook you can export to XLSX, with item analysis per question; quiz scoring starts on the Professional plan, and certificates for people who pass are on Business. See quizzes and assessments and Claform for education.

Frequently asked questions

What is a good pass mark for an online quiz?

Seventy percent is a sensible start for a training check, and 80 percent for compliance or safety topics, with critical questions that must be answered correctly. Check the guessing floor first: on a ten-question true-or-false quiz, someone who guesses every answer reaches 70 percent about one time in six.

How many questions should an online quiz have?

Five for a warm-up, 10 to 20 for a module check, and enough for a certification to cover each learning objective two or three times. Personality and product quizzes work best at six to ten questions, because nobody is obliged to finish them and every question should change the result.

How do I stop people from cheating on an online quiz?

Make guessing unprofitable with three or four plausible options, shuffle questions and options, give each person questions drawn from a larger pool, use personal access codes and hold back correct answers until all attempts are used. Proctoring signals help you spot unusual attempts, but treat them as a reason to ask, not as proof.

How many answer options should a multiple-choice question have?

Three or four, as long as every wrong option is believable. Michael Rodriguez's 2005 meta-analysis of multiple-choice research found that three options are usually enough. A fourth option that nobody picks makes the question longer to read without making it any harder to answer.

Start building for free