Test (online test)

Imagine two forms with questions. The first: "How would you rate the quality of our support?" — there is no correct answer here, only your opinion.

The second: "The capital of France is:

a) London, b) Berlin, c) Paris, d) Madrid" — here there is exactly one correct answer, and the system knows it.

The first form is a survey. The second is a test.

The difference seems obvious, but in practice the boundaries blur: tests are used for marketing, surveys are used to assess knowledge, and quizzes erase the line between entertainment and research altogether. Let's sort out what is what.

What a test is

Test — a structured form of checking knowledge, skills, abilities, or personal traits, in which each question (or group of questions) maps to a scoring system: correct and incorrect answers, points, result categories. On completion the participant receives a measurable result — a numeric score, a level, a type, or a recommendation.

If a survey asks "What do you think?" and accepts any answer as a given, a test asks "What do you know?" (or "Who are you?", "What are you capable of?") — and compares the answer against a built-in model. This is the fundamental distinction: a test has an internal scoring logic, a survey does not.

Testing is the process of taking a test. In the broad sense the word is also used outside questionnaire forms: software testing, A/B testing, load testing. In the context of this article we mean testing people — their knowledge, competencies, or psychological characteristics.

Test, survey, and quiz — drawing the line

The three formats overlap but serve different purposes.

A survey collects opinions. There are no correct answers. The goal is to learn what the audience thinks or does. The result belongs to the researcher: they analyze the data and make decisions.

A test assesses or classifies. There are correct answers (or there is a weighting model). The goal is to evaluate the participant. The result belongs to the participant: they learn their score, level, or type.

A quiz is a hybrid in which the test format is used for the sake of engagement. A quiz may check knowledge ("Guess the capital"), but more often it entertains and engages ("Which character from... are you?"). The main metric of a quiz is not the accuracy of the assessment but engagement: how many people completed it, shared the result, and clicked through to the site. More on that in a separate glossary article.

A survey is for the researcher. A test is for the participant. A quiz is for the marketer. In practice all three formats can live in one builder and even in one questionnaire — but understanding the differences helps you choose the right mechanics.

Types of tests

Tests are classified on several grounds. Here are the most practical ones — from the standpoint of the tasks solved through online builders.

Knowledge tests

The classic format: a question with options, one (or several) of which is correct. The outcome is the number of right answers, a percentage, a grade, or a level.

Where they are used:

• Education. Checking how well material was absorbed after a course, module, or lecture. A final exam or interim assessment.
• HR and corporate training. Employee appraisal, checking product knowledge, assessment after training. Especially in demand at companies with mandatory certification: banks, healthcare, industry.
• Recruiting. Preliminary screening of candidates: a test of knowledge of technologies, a language, or industry standards. Helps weed out unsuitable people before the interview stage.
• Marketing. "Check how well you know SEO" — a test as an entry point for attracting the target audience to the site.

Psychological and personality tests

Here there are no "correct" answers in the classic sense. Instead there is a weighting model: each answer option adds points to a certain category, and based on the total the participant lands in one of the types.

Where they are used:

• HR. Assessing motivation, work style, leadership qualities. Tests for detecting burnout, determining a personality type for team matching.
• Psychology and coaching. Diagnostic instruments: the Raven intelligence test, the Maslach burnout inventory, personality questionnaires.
• Entertainment and virality. "Which Harry Potter character are you?" — a result that the participant happily shares on social media. This is already closer to a quiz, but the technical foundation is a test with a weighting model.

Skills and competency tests

These check not theoretical knowledge but the ability to apply it in practice. They often include situational tasks, cases, and analysis assignments.

Example. A test for a sales manager: "The client says your product is too expensive. What do you reply?" — three options, each scored from the standpoint of sales technique. There is no correct answer in the absolute sense, but there are more and less effective strategies, and the system assigns each its own weight.

Certification tests

Formalized exams whose results determine whether the participant receives (or does not receive) a certificate, clearance, or license. The requirements for reliability and security are at their maximum here: a time limit, a ban on returning to previous questions, protection against cheating (randomization, a question bank).

Scoring mechanics: how the assessment works

Behind any test stands a numeric model. There are several ways to arrange that model.

Simple counting of correct answers. Each correct answer is 1 point, an incorrect one is 0. The result: "7 out of 10", "70%". The most transparent and understandable system. Suited to knowledge tests where all questions are equivalent.

Weighted scoring. Different questions are "worth" different amounts. A hard question is 3 points, an easy one is 1. This lets you reflect priorities: if, in a safety test, the question about first aid is more important than the question about where the fire extinguisher is, it should weigh more.

Categorical model. Answers are not scored as correct or incorrect — instead each option adds points to one of the categories. In the end the participant lands in the category with the highest score: "You are a visual learner", "Your leadership style is democratic", "You are a Gryffindor". This is exactly how personality tests and entertaining quizzes are built.

Threshold model. The result is determined not by an absolute score but by falling within a range: 0–40% — "Beginner", 41–70% — "Confident", 71–100% — "Expert". The thresholds are set by the test author, and they can be tied to specific actions: score below 60% and you are sent to retake the training.

How to build a good online test

Start with the goal, not with the questions. What do you want to measure? Product knowledge? Language level? Personality type? Everything else depends on the goal: the type of questions, the scoring system, the result format. A test created "just because" is as much a waste of resources as a survey without a goal.

Word the tasks unambiguously. In a knowledge test it is especially important that the correct answer be indisputably correct. If two experts cannot agree on which option is right, the question is not ready. Ambiguous wording is the main source of complaints from participants and the main killer of trust in the results.

Use randomization. Shuffling questions and answer options is the basic defense against cheating. If the order is fixed, participants quickly pass each other the scheme "in the third question the correct one is the second option". With randomization that hint stops working.

Balance the difficulty. A test in which all questions are elementary does not distinguish strong participants from weak ones — everyone gets 90–100%. A test where all questions are impossibly hard demoralizes and yields no useful data. The optimal structure: 20–30% easy questions (a warm-up, basic confidence), 50–60% medium ones (the bulk that determines the result), 10–20% hard ones (separating the good from the excellent).

Add feedback to the answers. After each question (or at the end of the test) show where the participant erred and why the correct answer is what it is. This turns the test from an assessment tool into a learning tool. Participants value this and perceive the test not as an exam but as a useful experience.

Set a reasonable time limit. For certification and exam tests a time limit is the standard. For training ones it is optional. If there is a timer, it should be realistic: give the participant enough time to think, but not so much that they can google every answer.

Common mistakes

Distractors that are too obvious. If in the question "Which programming language was created by Guido van Rossum?" the options are Python, Microwave Oven, Ketchup, Gardening, the test does not check knowledge — it insults the participant's intelligence. Incorrect options must be plausible — otherwise the question turns into a formality.

All correct answers are the longest. This is a surprisingly common mistake: the test author phrases the correct option more thoroughly and carefully than the incorrect ones. The participant quickly notices the pattern and starts choosing the longest option without reading the question. Make sure all options are roughly the same length and style.

No review of mistakes. The participant got "6 out of 10" — and then what? Without explanations of where and why they erred, the test does not teach, it only frustrates. This is especially critical in corporate training: an employee should not merely learn their score but understand what to brush up on.

No protection against retaking. If a test can be taken any number of times and the correct answers can be seen, its certification value is zero. For serious tests you need an attempt limit, a question bank (out of 50 questions each participant randomly gets 20), and time control.

Tests in SurveyNinja

The SurveyNinja builder supports the full cycle of creating tests — from a simple knowledge test to a personality quiz with categories.

Correct and incorrect answers. For each question you can mark the correct options and assign points. The system automatically tallies the result and shows it to the participant on the results screen. More in the "Correct/incorrect answers and scoring" guide.

Results screens by range. You can set up different final screens depending on the score: "0–40% — Beginner", "41–70% — Confident", "71–100% — Expert". Each screen has its own text, image, and recommendations.

Randomization of questions and answers. At the questionnaire level and at the level of each question — shuffling the order. Combined, this creates a unique version of the test for each participant, which is critical for preventing cheating.

Timer. A time limit on completion — configurable for the whole test or for individual pages. When the time runs out, the test finishes automatically with the current result.

Ready-made templates. The template library has dozens of ready-made tests: from a literacy test to the Raven test and a marketing knowledge test. You can use them as is or adapt them to your needs.

A test is not just "questions with correct answers". It is a measurement instrument, and its quality determines whether you get a reliable assessment or random numbers. A good test is unambiguous, balanced in difficulty, and useful for the participant — not only for its creator.