How to Make Online Quizzes People Finish: 10 Design Tactics

Good online quiz design starts with respect for the learner’s time. Tell people how long the quiz takes, keep every question earning its place, make it easy to answer on a phone, and check where people leave. This post is for trainers, instructional designers and L&D managers who build knowledge checks and course quizzes, not lead-generation quizzes for online stores.

Most advice on this topic rests on vendor numbers with no method behind them. Here, each tactic is labeled by what backs it: an experiment or dataset, official statistics, vendor guidance, or our editorial judgment. Most experiments studied surveys, not quizzes, so treat them as strong hints, not proof.

One more point before the tactics. A finished quiz is not always a good one. In a web experiment by Galesic and Bosnjak (2009), answers to later questions were faster, shorter and more uniform than answers near the start. The aim is people who finish and still read the questions.

Key takeaways

• State the length and purpose before the first question. When an experiment varied the stated length between 10, 20 and 30 minutes, longer stated lengths meant fewer people started and finished.
• Cut every question that would not change a decision, and split what remains if it has to stay long.
• Plan for phones. In six Dutch government surveys in 2017, smartphone break-off was 32.1%, against 9.6% on computers.
• Count exits per question, not just the overall completion rate. One weak item can cause most of the loss.

Which online quiz design tactics have evidence behind them?

Few tactics have direct evidence for quizzes. Some have survey experiments behind them, a few have official statistics, and the rest are vendor or practitioner guidance. The table labels each tactic so you can decide how much weight to give it. “Editorial” means our judgment, with no study found.

How do you set up a quiz before anyone starts?

Before the first question, decide three things: what the learner is told, how many questions you ask, and whether one long quiz should be several short ones. All three affect who starts and who finishes, and all three are cheap to change.

Tactic 1: Say how long it takes, why it matters and what the learner gets. In the Galesic and Bosnjak experiment, the stated length was 10, 20 or 30 minutes, and “the longer the stated length, the fewer respondents started and completed the questionnaire.” Put the question count, the minutes and one line on its purpose on the first screen. An illustrative welcome line: “8 questions, about 4 minutes. Your answers show which module to repeat.” Time a pilot group first so the promise is true.

Tactic 2: Cut to the decisions the quiz supports. SurveyMonkey analyzed 100,000 random surveys, 2,000 for each length from 1 to 50 questions. More questions meant more drop-off, and the steepest rise per added question came up to 15 questions. The surveys ran from January 2009 to September 2010, so use the direction, not the numbers. For each item, ask what you would do differently depending on the answer. If nothing changes, cut it.

Several quiz-maker help pages suggest roughly 5 to 12 questions, but they are written for marketing quizzes and publish no method. Do not copy that range into a pass/fail assessment: three to five items cannot support that decision alone, as the post on knowledge checks in live virtual sessions explains. Split such a quiz instead of cutting it.

Tactic 3: Split long quizzes into short parts. Andreadis and Kartsounidou (2020) tested a long questionnaire against a split version with 1,090 sampled people. Drop-outs were 7.5% for the long version and 3.6% for the split one. But the split version’s combined response rate was lower (20.46% against 34.5%), and its second part went out at least 15 days later. The authors did not recommend splitting yet.

For training, split with a short gap, not a long one. A 2013 PNAS study by Szpunar, Khan and Schacter (80 students, a 21-minute lecture video) put short tests between segments. The tested group reported mind wandering on 19% of probes, against 41% for the group with no tests.

What should the quiz do while people answer?

Make each screen easy to answer: an easy first question, one question at a time on phones, plain wording, honest progress cues and quick feedback. These choices keep the effort of each answer low, so the questions get a real attempt instead of a skim.

Tactic 4: Open with an easy question. Quiz-maker guides advise a simple first item before harder ones. None we read publishes an experiment, so this is editorial support. In a scored assessment, use an unscored warm-up item instead. Because later answers were faster and shorter in the Galesic and Bosnjak experiment, also randomize scored items where your blueprint allows, so one topic is not always answered last.

Tactic 5: Show one question per screen, built for thumbs. In Statistics Netherlands data (Roberts and Bakker, 2018), break-off in six 2017 surveys was 32.1% for smartphone users, 12.1% for tablets and 9.6% for computers. Those were official surveys, not quizzes. Still, test on a phone if your learners answer away from a desk. Use large tap targets, avoid wide grids and long typed answers, and keep images light.

Tactic 6: Show progress, but do not expect it to work alone. A meta-analysis of 32 randomized experiments by Villar, Callegaro and Yang (2013) found that “using a constant progress indicator does not significantly help reduce drop-offs.” Indicators that moved fast first and slowed later reduced drop-offs, while the reverse increased them. Our reading: a plain “Question 4 of 10” is fine for orientation, and short items early make progress feel quick. Never fake it.

Tactic 7: Write items that can be answered in one read. Practitioner lists from e-learning and library teams advise against negative stems (“Which should you not do?”), trick questions and implausible distractors, and suggest roughly three to five options. These are guidelines, not finish-rate findings. For choosing formats that fit each skill, see this guide to question types for professional tests.

Tactic 8: Give feedback on wrong answers when the quiz is for learning. Articulate’s community guidance recommends relevant feedback on incorrect answers so the quiz teaches as well as tests. For a scored exam, do the opposite: hold answers until the window closes for everyone, as the exam security settings checklist advises.

Pro tip: Run The Thumb Test. Open the quiz on a mid-range phone, on mobile data, and answer it standing up with one thumb. Note every item where you zoom, scroll sideways or mistype, and compare your time with the minutes you promised on the first screen.

How do you handle interruptions and find where people leave?

Let people stop and come back, then count how many reach each question. The overall completion rate tells you something is wrong. The per-question count tells you where.

Tactic 9: Let people pause and return. We found no study on this, so it is editorial. Learners will be interrupted during a working day. Save answers as they go, allow back navigation unless exam rules forbid it, and set how long a paused attempt stays valid.

Tactic 10: Track exits per question, not just the overall rate. Count how many learners reached each question. The exit rate at question n is the number who reached n minus the number who reached n+1, divided by the number who reached n. One quiz-maker help page also suggests that a time spike of two to three times the other questions signals confusion, with no source.

An illustrative example: 200 learners start a 10-question onboarding quiz and 126 submit, a 63% completion rate. Exit rates at nine of the ten items sit between 1.5% and 3.2%. At question 5, 180 learners arrived and 140 reached question 6, so 40 left, an exit rate of 22.2%. That one item caused 40 of the 74 exits. Fix question 5 first, change one thing, and compare the next cohort. For items people answer but get wrong in odd patterns, use item analysis for trainers.

The table below is our diagnostic guide, applying the per-question method above and the speeding finding from Galesic and Bosnjak. It is editorial, not a validated model.

How should online quiz design change for compulsory training?

When people must finish, completion rate stops being a useful signal, because compulsion hides weak design. Judge the quiz by time spent, how uniform the answers look, and whether results match what you know about the learners. The table shows what to lean on in each situation.

Pro tip: Report Median Time Beside Completion. For mandatory quizzes, show median time per item next to the completion rate. After a pilot, pick a cutoff such as half the median and flag attempts where most answers fall under it for a human look, not an automatic penalty.

What should you do next?

Pick one quiz that is already live. Pull the number of learners who reached each question, find the biggest single drop, and change one thing there. Run the next cohort and compare.

Then write the welcome line from real pilot timings and take the quiz yourself on a phone. If you are choosing a platform, SimpliAssess’s published features and FAQ pages list analytics and reporting, custom reports and charts, and support for mobile devices and tablets. They do not describe progress indicators, save-and-resume or per-question drop-off reports, so you can request a SimpliAssess demo and ask about those directly.

Frequently asked questions

How many questions should an online quiz have?

It depends on the decision the quiz supports. In SurveyMonkey’s 2009 to 2010 survey data, drop-off rose steepest up to 15 questions. Marketing quiz pages suggest 5 to 12 without published methods. For a knowledge check, cut anything that changes no decision. For pass/fail, keep enough items to support the call and split the quiz into parts.

What is a good quiz completion rate?

There is no reliable universal benchmark. Vendor pages quote figures from roughly 54% to 70% for marketing quizzes, mostly without methods or samples. Record your own baseline for the same quiz and audience, change one thing, and compare. For mandatory training, completion is close to 100% by design, so watch time per item instead.

Does a progress bar help people finish a quiz?

Not reliably. A meta-analysis of 32 randomized survey experiments found that a constant progress indicator did not significantly reduce drop-offs, and with small incentives it increased them. Indicators that moved fast early on helped. A simple “Question 4 of 10” is still useful for orientation, but do not count on it to retain people.

Is one question per page better than a scrolling quiz?

For phones, it is a sensible default. Quiz-maker guides recommend it, and official Dutch survey data show far higher break-off on smartphones than on computers. We found no experiment comparing the two formats for quizzes, so test both with your learners on their own devices before you commit.

Should a quiz show correct answers right away?

For learning checks, yes. Feedback on wrong answers helps the quiz teach as well as test. For a scored assessment, hold correct answers and feedback until the window closes for everyone, so early finishers cannot share them. Decide per quiz and write the rule down before launch.

Why do people quit a quiz halfway through?

Usually the quiz is longer than promised, one item is confusing or hard to answer on a phone, or fatigue sets in. Your per-question exit counts show which. Evenly spread exits point to length. Exits clustered at one item point to that item.