An auto-scored quiz is a form where the answers carry values. Set a point value per question, and every submission arrives graded: thirty students, a training cohort, or a self-assessment on a landing page, marked before you have opened the tab.
The mechanics take about five minutes. The part worth thinking about is what happens after the score.
Step 1: Decide what the score is for
There are three different quizzes hiding behind the same feature, and they want different settings:
A teaching check. The score barely matters; the per-question feedback is the whole product. Release everything immediately.
A record of competence. Compliance training, onboarding, certification practice. The score matters, so the questions have to be defensible and the mark should be released deliberately.
A marketing quiz. A scored self-assessment that produces a “you are a…” outcome. The score is a device; the interesting question is what you show at the end.
Pick one before you write questions. A quiz trying to do all three does none of them.
Step 2: Write questions with one idea each
A question that tests two things cannot be scored fairly, because a wrong answer does not tell you which of the two failed. If a question has an “and” in it, it is usually two questions.
Use closed-set types for anything you want graded automatically:
- Multiple choice for a single correct answer
- Checkboxes for “select all that apply”, where the point values give you partial credit
- Dropdown for matching against a long list
- Image hotspot for “label this part of the diagram”: genuinely better than a text question for anything spatial
- Priority matrix for ordering steps
Written answers are fine to include, but they need a human, and that changes when you can release the mark.
Step 3: Write the wrong-answer feedback first
This is the trick that separates a quiz that teaches from one that tests.
Before you write the explanation for the correct answer, write the explanation for each wrong one. It forces you to know why each distractor is tempting: and if you cannot articulate that, the distractor is not doing any work and should be replaced with one that is.
In Questionful, each question can carry different feedback for right and wrong answers, and that feedback can include a link or an embedded video. The explanation lands at the exact moment the respondent is wondering why, which is the only moment it is genuinely absorbed.
Step 4: Set the points
Attach a point value to each question. A few rules of thumb:
- Equal weight unless you have a reason. Unequal weighting needs to be defensible if anyone asks.
- Do not put ten points on a knowledge question and one on a reasoning question, and then claim to measure reasoning.
- For checkbox questions, remember partial credit is a consequence of your point values: decide whether you want it.
Closed-set questions are scored the moment the response arrives. There is no marking queue for them.
Step 5: Decide what the respondent sees
Four separate settings, and they matter more than people expect:
- When the mark is released: immediately, or after you have manually reviewed the written answers.
- Whether they see missed questions.
- Whether they see correct answers.
- Whether they see point values.
For a teaching check, show everything immediately. For a quiz you plan to reuse with the next cohort, showing the correct answers hands your question bank to the internet, hold that back.
If the quiz has any written answers, use “release after review”. Otherwise respondents get a partial score that looks like a final one, and you will spend the afternoon explaining it.
Step 6: Set the presentation
Three settings that change the experience more than any of the above:
Slideshow mode puts one question on screen at a time. On the phones respondents will actually use, this is not a nice-to-have.
A progress bar turns an unknown quantity into a known one, which measurably reduces abandonment.
Shuffled question order gives every respondent a different sequence. It is the cheapest anti-copying measure available, and it costs nothing.
What automatic scoring cannot do
Worth knowing before you commit a certification programme to it:
- No pass/fail threshold. You get the score; deciding what counts as a pass is yours.
- No certificates. Nothing is generated for the respondent to keep beyond their score.
- No timers. Neither per question nor for the whole quiz.
- No question pools. You cannot draw ten questions at random from a bank of fifty: shuffling reorders the questions you wrote, it does not sample from a larger set.
- No proctoring. No lockdown browser, no identity verification, no monitoring.
- Written answers need a human. As they should.
If your programme genuinely requires timed, proctored, certificated exams, this is not the tool yet. If it requires a knowledge check that marks itself and explains the answers, it is a good fit. The quiz maker feature page has the complete list.
Step 7: Read the results as a question about your questions
The most valuable output of an auto-scored quiz is not the scores. It is the one question everybody got wrong.
Nine times out of ten that is a problem with the question: ambiguous wording, two plausible answers, or material you never actually taught. Export the responses to CSV, look for the item with the lowest success rate, and fix the question before you draw any conclusions about the cohort.
The short version
- Decide whether the quiz teaches, records, or entertains.
- One idea per question.
- Write the wrong-answer feedback first.
- Equal point weighting unless you can defend otherwise.
- Hold the mark if there are written answers; hide correct answers if you will reuse the quiz.
- Slideshow, progress bar, shuffled order.
- The question everyone failed is usually the question’s fault.
The online quiz template has the scoring and feedback structure already in place, and the exams and assessments use case covers the wider workflow, from building the check to exporting the results.