You are building the final quiz, and the tool offers to draw questions at random from a bank. Great, except now you have to decide how many questions to write. Ten feels thin. Fifty feels like a week of your life. And somewhere in the back of your mind is the worry that a random draw might hand one learner an easy quiz and another a brutal one. How many questions should a question bank have, and how do you make random draws fair?

A question bank should usually hold at least two to three times as many questions as you draw for each learning objective, so retakes feel fresh and every learner gets a comparable test. Group the bank by objective and draw a fixed number from each group, rather than drawing at random from one big pool. That keeps every attempt balanced across what the course actually teaches. Then write items that test one idea clearly, with believable wrong answers, and match the question type to the skill you are measuring. Bank size matters, but question quality matters more.

What Is a Question Bank and Why Use Random Draws?

A question bank is a pool of questions that a quiz pulls from, instead of a fixed list that every learner sees in the same order. With random draws, each attempt gets a selection from the pool. That has two big benefits for you and your learners.

How Many Questions Should a Question Bank Have? – Writing Assessments and Random Draws That Measure Something - 3D Cutaway Animation

First, retakes become real practice. If a learner fails and the second attempt shows exactly the same ten questions, they can pass by remembering answers rather than understanding them. Second, it cuts down on answer sharing, because colleagues sitting next to each other are not looking at identical quizzes.

Random draws do have a cost. You have to write more questions, and you have to make sure they are equally fair. A bank with three easy questions and three tricky ones on the same objective can give two learners very different experiences of the "same" test.

How Many Questions Should a Question Bank Have?

The number depends on how many you draw, how many times learners can retake, and how much repetition you can live with. A rule of thumb that suits many courses is to write two to three times the number you draw per objective. Let's work through a made-up example so you can see why.

Say your course has 5 learning objectives and the final quiz draws 10 questions, 2 per objective. If you write 6 questions per objective, your bank is 30 questions. On any retake, each question the learner saw last time has a 2 in 6 chance of being drawn again, about 33%. Across 10 questions, that means a learner sees roughly 3 repeats per retake on average (10 x 2/6 = about 3.3).

Now try 10 questions per objective, a bank of 50. Each question has a 2 in 10 chance of coming back, so about 2 repeats per retake (10 x 2/10 = 2). That is better, but you wrote 20 more questions to save about one repeat. For a low-stakes knowledge check, 6 per objective is often plenty. For a certification-style assessment with several retakes allowed, the bigger bank earns its keep.

Questions per objective (draw 2)Bank size (5 objectives)Chance a question repeatsAverage repeats per retake
42050%5
63033%3.3
105020%2

How Do You Group Questions by Learning Objective?

Group questions by learning objective by creating one small pool for each objective and drawing a set number from each pool. This is the single best way to keep random quizzes fair. Without it, a learner might draw four questions on the objective they know best and none on the one they struggle with.

Start with objectives written as observable actions. Michigan Tech's guidance on observable, measurable learning objectives is a good reference: "identify a phishing message" can be tested, while "understand phishing" cannot. If you cannot think of a question that tests an objective, that is often a sign the objective needs rewriting.

You can also weight objectives that matter more. If one objective covers a safety-critical step, draw 3 questions from its pool instead of 2. Just make sure that pool is bigger too, so the extra draws do not raise the repeat rate.

Within each pool, aim for questions of similar difficulty. One quick check is to ask your subject-matter expert to rate each question as easy, medium or hard. If a pool has mostly easy questions and one very hard one, the learner who draws the hard one has a different test. Rewrite or move it.

What Makes a Good Multiple-Choice Question?

A good multiple-choice question has a clear stem that makes sense on its own, one correct answer, and wrong answers that are believable to someone who has not mastered the material. Berkeley Lab's guide to writing multiple-choice questions covers the basics well: write a self-contained stem, keep distractors plausible and parallel, and avoid "all of the above," "none of the above" and negative wording.

Here is a quick before and after. A weak item might read: "Which of the following is NOT something you should not do with a suspicious email?" The double negative tests reading, not knowledge. A stronger version: "You receive an email asking you to confirm your password through a link. What should you do first?" Then offer four realistic actions, such as clicking to check the link, replying to ask if it is real, reporting it with the reporting button, or forwarding it to a colleague.

The best distractors come from real mistakes. Ask your expert, "what do people actually do wrong here?" and use those answers. A distractor nobody would ever pick wastes a slot and makes the question easier than it looks.

Which Question Types Fit Which Skills?

Choose the question type that is closest to how the skill is used on the job. Multiple choice is flexible and easy to score, but it is not the only option, and some skills are measured better another way.

  • Multiple choice and true/false: recognizing facts, rules and correct actions.
  • Multiple response: situations where several things are true at once, such as spotting all the hazards in a scene.
  • Sequence: procedures where order matters.
  • Matching: connecting terms, roles or tools to their uses.
  • Hotspot: finding something in an image, such as the right field on a form.
  • Fill in the blank or text entry: recall of a short, exact answer, used sparingly since spelling variations cause trouble.

For judgment calls with consequences, a short scenario often beats a stack of single questions. And if you only need a quick check after one idea, a couple of questions inside short microlearning modules can do that job without a big final bank.

How Should Retries and Review Work?

Retries and review should help learners improve without letting them memorize their way to a pass. A common, balanced setup is a pass mark that reflects what the job needs, a limited number of retries with a fresh draw each time, and review that shows which objectives were missed.

Think about what review reveals. Showing the correct answer for every question right after a failed attempt is kind, but with a small bank it also hands out the answers for the next attempt. Showing which objectives the learner struggled with, with a link back to that part of the course, keeps the help while protecting the bank.

Once the quiz is live, let the results teach you about the bank. Many LMS reports can show how often each question is answered correctly. A question that almost everyone misses may be badly worded or may point at a gap in the course, and one that everyone gets right may be giving the answer away. Review those outliers with your expert after the first few weeks, fix or replace them, and keep a note of what changed so later results still make sense.

If you build in AuthorLearn Studio, question banks with random draw are built in, along with around twenty question types and a results slide with a pass mark, retry and review settings. Before you publish, a plain-words box tells you how a given score will be reported, which is a handy last check that your pass mark does what you intended.

A Question Bank Is Only as Fair as Its Weakest Question

A good question bank holds two to three times the questions you draw for each objective, is grouped so every attempt covers the whole course, and is written carefully enough that any random draw gives a fair test. The math tells you how big the bank needs to be. The writing decides whether it measures anything.

So write your objectives first, size each pool with the draw math, and get your expert to supply the real mistakes for your distractors. Then read every question as a learner would, and fix the one that makes you wince. That question is the one some unlucky person will draw.