What are the main survey question types?
Common survey question types include single choice, multiple selection, numeric entry, ratings, agreement items, ranking and open-ended questions. Each captures a different kind of information. Choose the format from the question you need to answer and how you will interpret the result.
A matrix is a way to display several items together, not necessarily a different measurement type. A date field is useful for timing. A slider changes the input experience but does not automatically make a rating more precise.
No type is always best. A useful survey can use one format or several. Open text can reveal an experience in the respondent’s own words; it is not mandatory after every number, nor does it automatically establish why an outcome occurred.
Choose the type that fits the job
Scroll horizontally to see all columns →
| Type | Useful for | Example | Analysis consideration |
|---|---|---|---|
| Single choice | One category or answer | Which program are you reviewing? | Options should fit the intended single choice |
| Multiple selection | Several applicable categories | Which services did you use? | Percentages may add to more than 100% |
| Numeric entry | A count or amount | How many sessions did you attend this month? | Define unit, period and valid values |
| Rating | Level, frequency or evaluation | How satisfied were you with registration? | Use clear labels and a suitable scale |
| Agreement item | Response to a statement | The instructions were easy to follow | Agreement is not the same as observed behavior |
| Ranking | Relative order | Rank these three workshop topics by priority | Ranks do not show the size of preference differences |
| Open-ended | Unanticipated detail or explanation | What would make registration easier? | Review and code before reporting theme counts |
These examples illustrate formats, not a validated questionnaire. Adapt the language and test whether the intended audience interprets each item as expected.
Single-choice questions
Use a single choice when one answer meaningfully describes the situation. For a reporting period, “Which program are you reviewing?” may have one intended answer. “Which programs did you attend?” may require multiple selection instead.
Check category boundaries. Ranges such as “1–5” and “5–10” overlap at five. If the categories do not cover a legitimate answer, a respondent may choose an inaccurate option simply to continue.
Include “not applicable,” “not sure” or another appropriate alternative when it represents a meaningful response. Do not treat those options as interchangeable with a neutral opinion.
Multiple-selection questions
Use multiple selection when several options can apply. State whether respondents should select all relevant options or a limited number. Those instructions produce different data.
In a fictional survey of 100 respondents, 60 select email updates and 50 select text messages. That is 60% and 50% of respondents. The total exceeds 100% because a person may select both. Dividing by the 110 selections would answer a different question: the share of selections.
Test “none of these” behavior if it is offered. Selecting it alongside a listed option may be contradictory. Decide how the form and analysis should handle that case.
Numeric and date questions
Specify the unit and reference period. “How many?” is incomplete without defining what is counted and when. Visits, unique people and attendance entries can produce very different numbers.
Choose sensible validation without rejecting legitimate values. A negative count is usually invalid, while a negative financial change may be meaningful. A high value should sometimes trigger clarification rather than automatic deletion.
Keep zero distinct from missing data. For dates, explain which event date you need and use an unambiguous format. Collection date, observation date and reporting period are not always the same.
Ratings and agreement items
A rating asks for a level or evaluation, such as satisfaction or frequency. An agreement item asks how strongly someone agrees with a statement. Use labels that fit the concept.
For example, “How often did you use the service?” needs frequency options. An agree–disagree scale would be an awkward way to answer it. “The instructions were easy to follow” can use agreement options, though a direct ease question may also be appropriate.
A single agreement item is often called a Likert-type item. A Likert scale commonly combines several related items. Do not create a validated-looking score by simply adding unrelated answers together.
Keep the order and meaning of response options understandable. If you report an average, explain the coding and consider whether the distribution conveys important differences hidden by the mean. The Likert scale guide develops these choices.
Ranking questions
Ranking asks people to put options in order. It is useful when the decision concerns relative priority among a manageable set of alternatives.
Keep the list short enough to make the task realistic. Ranking ten unfamiliar items may create effort without reliable preference information. Consider whether a top-choice question or separate ratings would better serve the decision.
A first-place rank does not mean the option is liked, and the difference between ranks one and two is not necessarily the same as between two and three. Someone may rank an option first while finding all options unhelpful.
Open-ended questions
Open questions allow detail you may not have anticipated. Ask a focused question rather than placing a large blank box after every item. “What made the first appointment difficult to arrange?” is more directed than “Any comments?”
Use neutral wording. If you do not know whether there was a difficulty, ask about the experience first or provide wording that does not presume a problem. Leave room for mixed and positive accounts.
Plan the review before collection. Themes should be supported by the responses, with ambiguous cases visible and relevant source passages available to authorized reviewers. A frequent theme describes what respondents discussed; it does not prove the cause of a score or outcome.
Read more about open-ended survey questions and closed-ended questions.
Use matrix layouts carefully
A matrix can present several items using the same response options. It can also become difficult to read on a phone or encourage people to move through rows without considering each one.
Test the actual layout on the devices respondents use. Keep row wording concise, avoid combining unrelated concepts and check whether separate questions would be clearer. A compact screen for the designer is not necessarily an easier task for the respondent.
Each row still needs a clear definition and analysis plan. Do not treat all matrix answers as one score unless the measurement approach justifies it.
Write the question before polishing the interface
AAPOR recommends specific, understandable wording, one concept at a time and carefully chosen response options. Its survey research guidance also discusses the tradeoffs between open and closed questions.
“Were the trainer and materials useful?” combines two judgments. Someone may value the trainer and dislike the materials. Split the question when the separate answers are needed for a decision.
Check the reference period, eligibility and wording with intended respondents. Use survey logic for relevant paths, while retaining the information needed to explain who was asked each question.
Keep answers connected to the question they answer
Store the question definition, response values, period and relevant context with the results. Answers from the same anonymous submission can be analyzed together without identifying the person. Individual follow-up requires a different, appropriately governed connection.
A rating and comment in the same row are not disconnected simply because they occupy different columns. The useful requirement is that the team can interpret them together, inspect the source and keep the question’s meaning intact across imports or reports.
For multi-site or network collection, agree the small common set of measures needed for comparison. Local questions can differ. Preserve incompatible definitions rather than forcing every numeric answer into one total.
If a question or scale changes, record the version and assess comparability. Changing from “select all” to “choose one” changes what the answer represents, even when the options stay the same.
Test one result before launching
Build a small pilot dataset containing typical answers, missing values, multiple selections and a difficult comment. Produce the chart or table you expect to use.
- Can you explain the denominator?
- Are not asked, unanswered and zero distinguishable where required?
- Does the format capture the decision-relevant information?
- Can a reviewer inspect the source behind a theme?
- Can the team maintain definitions and changes?
Sopact’s focus is collection, analysis and governance of continuing evidence. Evaluate the question types and controls needed for your workflow in the proposed configuration. A software category does not guarantee that every format or analysis is supported.
Continue with survey software requirements or survey analysis.
Frequently asked questions
Which question type is best?
The one that captures the information required for your decision with an understandable task. No single type is best for every survey.
Does every rating need a comment?
No. Add narrative where it provides useful context and you have a plan to review it. Avoid unnecessary burden.
Is a matrix a separate measurement type?
It is primarily a layout for several items, often sharing response options. Each item still needs a clear meaning and analysis plan.
Can multiple-choice percentages exceed 100%?
Yes, when respondents can select several options and percentages use respondents as the denominator. State that clearly.
Does a ranking show how strongly someone prefers an option?
It shows order, not the size of the difference between preferences. Use another measure if strength of preference matters.
Do answers need a personal identifier to be analyzed together?
No. Answers within an anonymous submission can remain connected. Personal tracking is only needed when the intended design calls for it and the collection arrangement supports it.
