A quiz that marks itself is a different object from a form with right answers in a spreadsheet afterwards. The marking is what makes it worth taking, and the result screen is what makes it spread.
The mechanics are straightforward. The decisions around them are where quizzes go wrong.
The mechanism, in four parts
Every self-marking quiz is the same four things, whether it is a classroom test or a personality quiz.
- A variable that holds the running score. It starts at zero.
- Points on answers. Each correct option adds to the variable. For a personality quiz, different options add to different variables.
- A rule per outcome. "If score is 8 or above, go to this ending." Evaluated in order, so the catch-all goes last.
- An ending screen per outcome, with the result and what to do next.
Deciding what counts as a good score
Set the bands before you write the questions, not after you see the results. Deciding the pass mark once you know that most people got six out of ten is how a quiz ends up measuring nothing in particular.
Three or four bands works for knowledge quizzes. Two is a pass or fail, which is right for compliance and demotivating for anything else. More than five and the differences stop meaning anything to the person reading them.
For a personality quiz, four to six outcomes. The constraint is different: each one has to be distinct enough that somebody recognises themselves, and flattering enough that they share it. Nobody posts a result that reads as an insult.
Writing questions that score cleanly
The wrong answers matter more than the right one. A question with three obviously absurd options and one correct answer tests nothing. Each distractor should be a mistake somebody actually makes, which means the wrong answers tell you which misconception a person holds.
Avoid questions where two answers are defensible. If you find yourself writing "the best answer is", the question needs rewriting, because somebody will argue and they will be right.
And keep the scoring flat unless you have a reason. Weighting one question at triple feels precise and mostly produces results nobody can explain.
How long
Seven to twelve questions for something people take voluntarily. Completion falls off sharply past fifteen, and a quiz nobody finishes scores nobody.
Eight to twelve for a classroom quiz, so there is time to discuss it in the same lesson. Compliance checks can be longer because attendance is not optional, but the attention still is.
Where to ask for an email
After the last question and before the result. Never before the first.
Asking up front costs you most of the people who would have finished, because you are charging admission to something they have not yet decided they want. Asking at the end works because somebody who has just spent two minutes answering has invested something, and the result is a reward they can see.
Make it optional and show the result either way if you can bear to. The conversion difference is smaller than people expect and the goodwill difference is not.
Asking for an email before the first question charges admission to something nobody has decided they want yet.
The result screen is the product
A number on its own is a dead end. What makes a result worth reading is the interpretation: what this score means, what it suggests about them, and what to do next.
Show which questions they got wrong, if it is a knowledge quiz. It is the only part that teaches anything, and it is the reason people retake.
Give each outcome its own next step. A top score and a poor one should lead somewhere different, and for a lead qualification quiz that difference is the entire point: a strong fit gets a booking link, a poor fit gets something genuinely useful instead of a sales call neither party wanted.
Three mistakes worth avoiding
These account for most quizzes that get built and then quietly abandoned.
- Scoring decided after the fact. Set the bands first or the result means whatever makes the numbers look reasonable.
- One outcome everybody gets. If eighty percent land in the same band, the bands are wrong, not the respondents.
- A result with no next step. The most engaged a visitor will ever be is the second after they see their score, and most quizzes spend it on a full stop.
Quiz formats that work, and one that does not
Not every question type scores cleanly, and the ones that do not cause most of the arguments.
- Multiple choice with one correct answer. The default. Scores unambiguously.
- True or false. Fast, and worth half the points of a four-option question because guessing gets you fifty percent.
- Select all that apply. Decide in advance whether partial credit counts. Both are defensible; silence is not.
- Put these in order. Good for process questions and genuinely hard past about five items.
- Estimate a number. Score by how close they get. Underused and excellent for anything quantitative.
- Free text. Does not score automatically and never will reliably. Use it as an unscored closer, not as a question.
Where quizzes actually get used
Four quite different jobs, with different constraints.
- Lead generation. A personality or recommendation quiz where the result is the incentive. Email after the last question, never before.
- Qualification. Scored against fit, routing strong matches to a booking and everybody else somewhere useful.
- Teaching. Self-marking, with the per-question breakdown showing what the class did not understand.
- Compliance. A pass mark, a record of who took it and when, and a route to retake it.