Question selection
10 types of survey questions: examples and when to use them
Choose the right survey question type with 10 examples, a decision table, and a short message-testing questionnaire you can adapt.
The right survey question type depends on what you need to learn. Open questions reveal a person’s own words; single-choice questions classify one answer; multiple-response questions allow several; ordered ratings measure degrees of judgment. Choose the format after defining the decision, then check what information the format may hide.
Question types versus types of surveys
An online survey, telephone survey and face-to-face survey describe ways of collecting responses. A multiple-choice question, a ranking question and an open text question describe what the respondent does inside the instrument.
Mixing these categories can lead to the wrong planning decision. Moving a long matrix onto a phone screen changes the experience, but it does not turn the matrix into an open question. Plan the measurement and the collection setting together.
The table below is an editorial selection aid. It does not prescribe a fixed questionnaire or imply that all ten formats belong in one survey.
Choose from these 10 question types
Each example refers to a fictional service or a task that you would define before asking the question. The “watch for” column is part of the choice, because an easy-to-chart answer may be a poor measurement of the decision.
| Type | Example and useful use | Watch for |
|---|---|---|
| 1. Open text | In your own words, what does this service do? Useful for unaided interpretation. | Requires documented coding; writing effort varies. |
| 2. Single choice | Which plan did you use most recently? Select one. Useful for mutually exclusive classification. | Missing or overlapping categories. |
| 3. Multiple response | Which features did you use last week? Select all that apply. Useful for a list of reported activities. | Selections are not independent people; omissions may not mean no. |
| 4. Yes/no with routing | Did you contact support in the last 30 days? Useful for eligibility for a follow-up. | A binary choice may hide uncertainty or no recall. |
| 5. Agreement item | I can identify the renewal date. Useful for agreement with one proposition. | Agreement is not proof of understanding. |
| 6. Direct ordered rating | How easy or difficult was the task? Useful for a degree of judgment. | Labels and endpoints must match the construct. |
| 7. Numeric entry | How many times did you use the tool in the last seven days? Useful for a count. | Recall error, units and impossible values. |
| 8. Ranking | Rank these three priorities for the next update. Useful for relative order. | Does not measure how far apart preferences are. |
| 9. Paired preference | Which of these two headlines, if either, do you prefer? Useful for this explicit comparison. | Preference depends on the alternatives shown. |
| 10. Matrix | Rate each of three support tasks using the same difficulty options. Useful for repeated comparable items. | Mobile burden and careless repeated selections. |
A short message-testing sequence
Suppose a team needs to decide whether its headline explains a fictional budgeting app. The decision is about understanding the message, not whether the team can collect a high satisfaction score. An economical instrument might use four core steps.
First, show the message under the planned exposure conditions. Ask an open question: “What would this app help someone do?” Next, ask what information is missing. Then ask a direct relevance rating with a cannot-judge option. Finally, ask what influenced that rating.
If you show a feature list before the open answer, the later response is no longer an unaided restatement of the headline alone. Keep that distinction visible in the research plan. Pew’s discussion of order effects explains why earlier questions can shape later responses.
Sources: Pew on question order
Three decisions that often need a different format
To learn why people hesitate, do not start with a list of the team’s favorite explanations and assume it covers every barrier. An open response can reveal missing categories. Later, a carefully constructed closed question may help measure known reasons in a larger study.
To learn actual usage, ask about an explicit event and time period when the respondent can reasonably recall it. “I am a frequent user” measures agreement with a vague description. It is not the same as a reported count of sessions.
To compare claims, separate what a person understood from whether they liked it. A claim can be appealing because it was misunderstood. One combined score conceals that risk instead of resolving it.
Response options are part of the question
A single-choice item should make the intended categories clear. Consider a fictional monthly spend question with bands “$0–$50” and “$50–$100.” Someone who spent exactly $50 fits both. Non-overlapping boundaries, a currency and a time period are needed.
For ratings, keep a neutral judgment separate from lack of experience. For lists, decide whether none, other or cannot recall belongs. If the answer list does not allow an honest response, improving the sentence alone is insufficient.
Use a short pretest to find terms or categories the audience interprets differently from the team. The Census standard supports testing instruments and documenting their development. Your pretest should reveal edits to make, not manufacture a certificate of validity.
How AI can help choose a question type
Ask the assistant for a decision map before asking for 20 questions. For each proposed question, require the decision, the construct, the response format, and one interpretation risk. Remove rows where the answer would not change an action.
A useful prompt is: “Given this research decision, propose the fewest questions needed. For each, explain what it measures and what it cannot establish. Include routing, response options and an uncertainty choice where relevant. Identify questions that could reveal answers to later questions.”
Review the proposed sequence as a complete experience. An AI-generated questionnaire can be internally repetitive, contain inconsistent scales, or ask about experiences a respondent never had. The survey question checker supplies a first review pass and a prompt you can adapt.
What should appear in the report?
Keep the exact questions, response options, order and routing with the results. Include the analysis base for each percentage. A ranking, an agreement share and a feature-selection count are different kinds of data; do not present them as interchangeable evidence of demand.
If the survey used a sample recruited for convenience, the question format does not make that sample representative. Match conclusions to the population reached and the design actually used. Statistics Canada’s survey-evaluation questions provide a useful framework for checking these wider limits.
Sources and scope
- Writing Survey QuestionsPew Research Center
Methods guidance on wording, response options and order. The examples and review workflow here are original MTB applications.
- Statistical Quality Standard A2U.S. Census Bureau
Primary standards for developing and pretesting data-collection instruments; not certification of these templates.
- Questions to ask about surveysStatistics Canada
Questions for evaluating survey populations, designs and interpretation.
Sources support the specific statements described above; they do not validate this publication’s rubric, guarantee a compliant execution, or replace context-specific professional advice.
What this page is: a research-methods guide, not a report of completed consumer fieldwork. Corrections and material revisions are recorded under the publication’s editorial standards.