Survey Question Types (With Examples)

Pick one: single-choice questions

The workhorse. Use it when the options are genuinely mutually exclusive and you want a clean split in analysis. The failure mode is an option list that is not exhaustive — every single-choice question needs either a complete list or an escape option, otherwise people pick the least-wrong answer and quietly corrupt your data.

• How satisfied are you with your purchase?

  Single choice
  Very satisfiedSatisfiedNeither satisfied nor dissatisfiedDissatisfiedVery dissatisfied

  A balanced five-point scale: equal numbers of positive and negative options with a true midpoint. Unbalanced scales inflate positive results.

• Did you find what you were looking for?

  Single choice
  YesPartlyNo

  "Partly" earns its place here. A forced yes/no on a question with a real middle pushes people into whichever answer feels less wrong.

• How often do you use the product?

  Single choice
  DailyA few times a weekA few times a monthLess oftenThis was my first time

  Concrete frequencies, not "often" and "rarely" — vague frequency words mean wildly different things to different respondents.

• Which plan are you currently on?

  Single choice
  FreePaidNot sure

  Include "not sure". A surprising share of respondents genuinely do not know, and forcing a guess is worse than recording the uncertainty.

• Which best describes your role?

  Single choice
  Founder or ownerMarketingSalesProduct or engineeringOperationsResearchSomething else

Pick many: select-all questions

Use these when answers genuinely stack. The trade-off is that select-all data is harder to chart — each option becomes its own yes/no variable, and percentages no longer sum to a hundred. Say "select all that apply" in the question text itself, because the widget alone does not reliably communicate it.

• Which features do you use? Select all that apply.

  Select all that apply
  Form builderTemplatesAnalyticsIntegrationsExportsNone of these

  An explicit "none of these" option distinguishes a real negative from someone who skipped the question.

• Where did you hear about us? Select all that apply.

  Select all that apply
  Search engineSocial mediaA friend or colleagueA newsletter or blogAn adSomewhere else

  Self-reported attribution is unreliable but still useful directionally. Do not reconcile it against your analytics and expect the numbers to agree.

• Which problems have you run into? Select all that apply.

  Select all that apply
  Hard to get startedMissing a feature I needSomething was slowSomething did not workPricingNo problems

• Which channels do you prefer for support? Select all that apply.

  Select all that apply
  EmailLive chatHelp articlesPhoneCommunity forum

• Which devices do you use most? Select all that apply.

  Select all that apply
  Laptop or desktopPhoneTablet

Rating scales, Likert, and NPS

Scales turn opinions into numbers you can track over time, which is their whole value — a single satisfaction score means little, the same score across four quarters means something. Keep the scale identical between waves. Changing the number of points, the labels, or the direction breaks comparability with everything you collected before.

• On a scale of 1 to 5, how easy was it to get started?

  Single choice
  1 - Very difficult23 - Neutral45 - Very easy

  Label at least the endpoints and the midpoint. A bare row of numbers means each respondent invents their own interpretation of a 4.

• How likely are you to recommend us to a friend or colleague?

  Single choice
  012345678910

  The Net Promoter Score question. Scored by subtracting the share answering 0-6 from the share answering 9-10 — which means it discards the 7s and 8s entirely, so a raw average of this question is not an NPS.

• How much do you agree: the product does what I expected it to.

  Single choice
  Strongly agreeAgreeNeither agree nor disagreeDisagreeStrongly disagree

  A Likert item. Watch for acquiescence bias — people tend to agree with statements — and mix in some reverse-worded items if you are running several in a row.

• How would you rate the value for money?

  Single choice
  1 - Poor23 - Fair45 - Excellent

• How often does the product do what you need?

  Single choice
  AlwaysUsuallySometimesRarelyNever

Open-ended questions

Open questions are where you learn things you did not think to ask about, and they are the only question type that can tell you your survey was wrong. They cost the respondent real effort, so ration them: two or three per survey, placed after the closed questions, and never as the opening question.

• What were you trying to do when you first signed up?

  Long text

  Asks about a specific past event rather than a general opinion, which is what makes it answerable. "What do you think of us?" gets you nothing.

• What almost stopped you from buying?

  Long text

  The highest-yield open question in most commercial surveys. It gives people permission to be negative about something specific.

• If you could change one thing, what would it be?

  Long text

  "One thing" does the work — an unbounded request for feedback returns either nothing or an unusable list.

• What is your job title?

  Short text

  A short open field beats a role dropdown when you are segmenting professionals: messier to code, far more informative.

• What should we have asked you that we did not?

  Long text

  Put it last. It is the cheapest insurance against a badly designed survey, and the answers are often the most useful thing you collect.

Try it live

The same questions, asked one at a time

This is a working form built from the questions above — answer it the way a respondent would, then take it into your own account with one click.

See how conversational forms ask these one at a time →

Frequently asked questions

What are the main types of survey questions?

Four families cover almost everything: single-choice, where the respondent picks one option; select-all, where answers stack; scales, including rating, Likert agreement, and NPS; and open-ended, where they write freely. Everything else is a variation — dropdowns are single-choice with a different widget, matrices are several scale questions in a grid.

Should a rating scale have a midpoint?

Usually yes. Removing the midpoint forces people who genuinely have no strong view to pick a side, which manufactures opinion that is not there. Drop it only when you deliberately want to push respondents off the fence and you are willing to accept that some of the resulting answers are arbitrary.

How many open-ended questions should a survey have?

Two or three, placed after the closed questions. Open questions cost far more effort to answer than a tap on a scale, and a survey that opens with a text box loses people immediately. The one exception worth keeping is a final "anything we missed?" — it earns its place more often than not.