Wavefield Research
Sign in Start a project
Question craft

Closed-ended questions: examples of all seven types

By the Wavefield Research team · Published Sep 10, 2026

A closed-ended question is answered from fixed options — Yes/No, a choice list, a scale — instead of a text box. Closed questions produce countable, comparable data, which is why professional questionnaires are built mostly from them, and why choosing the right closed format decides what your analysis can honestly claim.

Below is a closed-ended questionnaire sample for each of the seven types, with answer options included, the analysis each format supports, the four bias traps, and the question the guides skip: when a question should be closed at all.

The types

Seven closed question types, with examples and their analysis

Each card shows example questions with real option sets, plus the one thing most guides omit — what the format yields statistically. Types with their own deep example banks link to them.

Dichotomous (yes/no)

  • Have you shopped with us before today?Yes / No
  • Do you favor or oppose the proposed bylaw?Favor / Oppose

Analysis: Yields a single proportion with the cleanest error math in research — and the yes-rate of a screener binary is your incidence, which prices your fieldwork.

Deep dive: 40 dichotomous examples →

Multiple choice, single select

  • Which one factor mattered most in your choice?Price · Quality · Speed of delivery · Brand · Recommendation · Other (please specify)
  • How do you usually listen to music?Streaming subscription · Free streaming with ads · Radio · Owned library · I rarely listen

Analysis: Shares of the answered base summing to 100%, comparable across segments with significance tests. The option list is the craft: exhaustive, exclusive, five to seven options.

Deep dive: 30 multiple choice examples →

Multiple choice, multi select

  • Which of these have you done in the past week? (Select all that apply)Ordered delivery · Cooked from scratch · Eaten out · Used a meal kit · None of these (exclusive)
  • Where do you get news about your industry? (Select all that apply)Newsletters · LinkedIn · Podcasts · Trade press · Colleagues · Other (please specify)

Analysis: The base is respondents, not selections, so columns legitimately sum past 100% — report % who selected, never % of selections. None-of-these must clear the rest mechanically.

Likert (agreement) scale

  • “The pricing page told me everything I needed to know.”Strongly disagree · Disagree · Neither · Agree · Strongly agree
  • “This company understands businesses like mine.”Same 5-point agreement scale — batteries share one scale so rows stay comparable

Analysis: Ordinal, not interval: distributions and top-2-box shares are always defensible; means work for comparison between segments or waves, never as absolute claims.

Deep dive: the full Likert guide →

Rating scale (including NPS)

  • How likely are you to recommend us to a friend or colleague?0–10, endpoints labeled Not at all likely / Extremely likely
  • How would you rate the value for money?1–5, all points labeled from Very poor to Excellent

Analysis: Numeric scales invite means, and 0–10 formats come with conventions attached — NPS nets promoters minus detractors, and its margin of error is wider than either share's alone.

Deep dive: the NPS calculator with margins →

Ranking

  • Rank these four improvements from most to least valuable to you.Faster support · Lower price · More integrations · Better reporting (drag to order)

Analysis: Forces trade-offs that ratings let respondents dodge. Analysis reports average rank or % ranked first; past six items, respondents fatigue and the tail ranks turn random.

Matrix / grid

  • How satisfied are you with each of the following?Rows: Support · Billing · Onboarding · Documentation — one shared 5-point satisfaction scale

Analysis: Efficient and the top source of straightlining — one column ridden all the way down. Rotate rows, cap the grid near seven, and let quality flags mark the straightliners.

The traps

Four ways closed questions quietly bias answers

Acquiescence: agreement is the path of least resistance

Respondents lean toward agreeing with whatever is asked. The fix costs nothing: name both sides in the question — favor or oppose, satisfied or dissatisfied — instead of asking people to endorse a one-sided statement. Balanced forced-choice wording is the standard professional correction.

Option order: first options win by being first

Early options get over-picked, on screens and especially on phones. Rotate the substantive options so position bias averages out across respondents, and anchor the conventional ones — Other, None of these, Prefer not to say — at the bottom where respondents expect them.

The false binary: closing a question that has a middle

A closed question can only return the options you wrote. If a real share of respondents sits between your options — undecided, mixed, it depends — a forced choice manufactures certainty. Either add the honest middle option or ask the binary and probe the leaners separately.

The invisible option: lists that miss the real answer

The most expensive failure is the option you never wrote, because it produces no error message — respondents pick the nearest wrong answer and the data looks clean. An Other option with a specify box is the alarm system: when it exceeds roughly 10%, your list is wrong.

None of these traps are exotic — Pew Research Center's questionnaire guidance documents all of them, and every one has a mechanical fix a platform can enforce: rotation with anchored escapes, exclusive None options, specify boxes on Other. On Wavefield the agent programs those defaults from the brief, which turns bias avoidance from a review checklist into the starting state.

The real decision

Close the question only after you know the options

The closed-versus-open decision is usually framed as a style choice. It is actually a knowledge claim: a closed question asserts that you already know the answer space. When that claim is true, closing is pure gain — faster answers, higher completion, countable results. When it is false, the closed format returns your own guesses with percentages attached, and the real answer hides inside an inflated Other or never appears at all.

The professional sequence resolves it: run the question open-ended when the territory is new, code the verbatims into categories, then field the closed version built from those categories when you need to measure at scale. Discovery pays the coding cost once; measurement collects the speed and comparability forever after. A tracker that skipped the discovery step is the most common way option lists end up wrong for years.

The economics of that sequence have shifted, which makes skipping it less excusable than it used to be. Coding open-ended verbatims was once the expensive step — analyst days per thousand responses — and closing questions early was a defensible cost saving. With AI-assisted coding, verbatims become countable themes in minutes, so the discovery wave now costs little more than the fielding itself. The asymmetry that remains is the one that always mattered: a coding pass can be redone, but data collected against the wrong option list cannot be recollected.

Where each format sits in the bigger toolbox is mapped in survey question types, and what happens to closed answers after fielding — weighting, crosstabs, significance — is the subject of how to analyze survey results.

FAQ

Common questions

What is a closed-ended question, with examples?

A question answered from fixed options rather than free text. Examples across the types: “Have you shopped with us before? Yes/No” (dichotomous); “Which one factor mattered most — price, quality, speed, brand?” (single select); “Rate your agreement: the checkout was easy — Strongly disagree to Strongly agree” (Likert); “How likely are you to recommend us, 0–10?” (rating). Closed questions produce countable, comparable data, which is why professional questionnaires are mostly closed with a few deliberate open-ends.

What are the types of closed-ended questions?

Seven cover practice: dichotomous (two options), single-select multiple choice, multi-select multiple choice, Likert agreement scales, rating scales (including NPS), ranking, and matrix grids. Dropdowns and sliders are presentation variants of these, not separate types — a dropdown is a single select that hides its options, which is usually a reason not to use one.

When should I use open-ended instead of closed-ended questions?

Use open-ends when you do not yet know the answer space — discovery, vocabulary, the why behind a number — and closed questions when you know the options and need to count them. The professional sequence runs open first: an open-ended question in an early study surfaces the real categories, coding turns them into an option list, and the closed version then measures at scale. Closing a question before you know its options bakes your guesses into the data.

Are closed-ended questions quantitative?

Yes — every closed format produces countable data, but not the same kind. Single selects produce shares that sum to 100%; multi selects produce per-option shares on respondent bases; Likert and rating items produce ordinal distributions where means need care; rankings produce order statistics. The analysis a question can support is decided the moment you pick its format, which is the real reason format choice matters more than wording polish.

Related: survey question types · dichotomous survey questions · multiple choice survey questions · 50 open-ended survey questions

Closed questions that behave, programmed for you

Brief the agent and it picks the right closed format for each question, rotates options with anchors, and reports what each format honestly supports. From $99 per study.

Start a study
Wavefield Research
support@wavefieldresearch.com Resources About Security Privacy Terms © 2026 Wavefield Research Inc.