Back
Survey Questions: Types, Examples and How to Write Them
SmartInterview Team
The Short Answer
Survey questions come in two families. Closed questions give respondents a fixed set of answers and produce countable data. Open questions let them answer in their own words and produce explanations. A good survey uses mostly closed questions to measure and a small number of open ones to understand, and every question earns its place by answering "what decision does this change?"
One idea per question. If the sentence contains "and", you probably have two questions.
No leading language. "How much did you enjoy our fast delivery?" is not a question, it is a press release.
Plain words. Write for someone reading on a phone while distracted, because that is who is answering.
Answer options must be exhaustive and mutually exclusive. Every respondent needs exactly one place to land.
Shorter is better. Drop-off compounds with each screen, and the people who quit first are often the ones with the most to say.
Skip ahead to the 25 examples if you just need wording you can copy, or start with the writing rules if your last survey came back unusable.
The Question Types, Compared
Each type answers a different kind of question and fails in a different way. This table is the map; the deep guides linked from each row are the territory.
Type | What it measures | Best for | Main weakness |
|---|---|---|---|
Single select | One choice from a list | Mutually exclusive facts: plan type, primary reason, main channel | Forces a single answer where reality is often plural |
Multiple choice (multi-select) | Any number of choices from a list | Awareness, usage, features owned | Later options get picked less; respondents satisfice and stop reading |
Dichotomous | A binary: yes/no, true/false | Screeners, eligibility, clean routing | Collapses nuance; wrong whenever "it depends" is the honest answer |
Likert scale | Degree of agreement or satisfaction | Attitudes, tracked sentiment | Scale use varies by culture and person; agreement bias |
Rating and NPS | Intensity on a numeric scale | Trend lines and benchmarks | A number with no reason attached is not actionable |
Rank order | Relative priority across items | Forcing trade-offs when everything rates as important | Cognitively heavy past 5-6 items; no sense of distance between ranks |
Matrix (grid) | Several items on one shared scale | Compact attribute batteries | Straightlining and mobile breakage; the most abused type in surveys |
Open ended | The respondent's own words | Reasons, unknowns, vocabulary | Needs coding to become countable; thin answers if the prompt is generic |
Go deeper on any of them: single select questions, multiple choice questions, dichotomous questions, Likert scales, rank order questions, matrix questions, open ended questions and closed ended questions.
The Anatomy of a Good Survey Question
A survey question has four parts, and a bad question is usually a failure in one specific part rather than a general lack of quality.
The stem. The question itself. One idea, plainly stated.
The reference frame. The time period and the object: "in the last 30 days", "for your most recent order". Missing frames are the most common cause of noisy data, because every respondent silently picks their own.
The response options. Exhaustive, mutually exclusive, and in an order that does not itself suggest an answer.
The escape hatch. "Don't know", "not applicable", "none of these". Without one, people who genuinely do not fit will pick something at random, and that noise is indistinguishable from signal.
Rules for Writing Good Survey Questions
One idea per question
Double-barreled questions cannot be interpreted. "How satisfied are you with our price and quality?" produces one number from two judgments, and you will never know which one drove it. Split it. The tell is the word "and" in the stem, though "or" and commas hide the same problem.
Remove leading language
A leading question tells the respondent which answer you want. Some are obvious: "How much did you enjoy our new design?" Some are subtler. Adjectives inside the stem ("our improved checkout"), asymmetric scales that offer three positive options and one negative, and agree/disagree framing all nudge answers upward. Neutral phrasing sounds flat. Flat is correct.
Use plain words
Cut internal vocabulary. Your customers do not know what your onboarding flow, your tiers or your modules are called. Cut negations, because "which of these do you not use?" is misread constantly. Cut double negatives entirely. Aim for the reading level of a news headline.
Do not assume knowledge or behavior
"How often do you use the reporting dashboard?" assumes they know it exists and have used it. Gate behavioral questions behind a screener, or include "I have not used this" as an option. Assumed-behavior questions manufacture usage data out of nothing.
Make options exhaustive and mutually exclusive
Age bands that read 18-25, 25-35 leave the 25-year-old choosing arbitrarily. A channel list without "other" forces a wrong answer from anyone unusual. If you use "other, please specify", read those responses: they are your list for next wave.
Watch order effects
Options at the top of a long list get selected more often on screen, and items at the end are favored when read aloud. For lists longer than about five items with no natural order, randomize. Keep scales and ordered categories in their logical sequence, always.
Do not ask what you already know
Purchase date, plan, region, order value: all sitting in your own systems. Passing them in as metadata rather than asking shortens the survey and avoids the irritation of being asked something the sender obviously knows.
Ask about behavior in the recent past, not in the future
Stated future intent is weakly related to what people actually do, especially on price and on "would you use this?". Anchor to something that already happened: what they last bought, what they did the last time this came up, what they are using today.
Question Order and Flow
A survey is a conversation with a shape, and the order changes the answers.
The standard structure
Screeners first. Disqualify fast, before anyone invests effort. Do not signal what qualifies.
An easy opener. One question anyone in scope can answer immediately. Never open with a grid, and never open with demographics.
Broad before narrow. Unaided awareness before you show any brand list, general satisfaction before attribute detail. Once you show the list, you cannot un-show it.
The core section. Whatever the study exists to measure, placed while attention is still high.
Sensitive questions late. Income, health, anything embarrassing. By then you have earned some trust, and an abandonment costs you less.
Demographics last. They are easy, and easy questions are what you want at the point where attention has gone.
Priming and contamination
Every question changes the questions after it. Ask about sustainability and the next satisfaction rating shifts. Ask about problems and the recommendation score drops. Neither is respondent dishonesty, it is how attention works. The practical defenses: put unaided questions before aided ones, keep contaminating topics after the measures they would affect, and hold the order fixed between waves of a tracker so at least the bias is constant.
How Many Questions Should a Survey Have?
Fewer than you have drafted. Two rules do most of the work.
Every question must map to a decision. Write the question, then write the sentence you will put in the report, then name the person who acts on it. If any of the three is missing, cut the question. This one exercise typically removes a third of a draft.
Length is measured in minutes, not questions. Ten open questions is a long survey. Thirty single-selects can be a short one. What respondents feel is effort and elapsed time. A short customer feedback survey lives at 3-6 questions; a substantial market research study runs longer, but past roughly ten minutes you should expect data quality to degrade as people speed up, straightline and stop reading. Falling response rates make this worse every year, as we covered in why survey response rates are crashing.
25 Survey Question Examples You Can Copy
Grouped by use case. Adapt the reference frame and the product noun; keep the structure.
Customer satisfaction and experience
Overall, how satisfied were you with your most recent order? (Very dissatisfied to very satisfied, 5 points)
How easy was it to get your issue resolved today? (Very difficult to very easy, 7 points)
How likely are you to recommend us to a friend or colleague? (0 to 10)
You gave that a [score]. What is the main reason for that number? (Open)
What is the one thing that would have made this a 10? (Open)
Did this experience match what you expected before you started? (Better than expected / About what I expected / Worse than expected)
Product and feature research
Which of these have you used in the last 30 days? (Multi-select, randomized, with "none of these")
Which one of these do you use most often? (Single select, piped from the previous answer)
Rank these five capabilities from most to least useful for your work. (Rank order)
How disappointed would you be if you could no longer use this product? (Very disappointed / Somewhat disappointed / Not disappointed)
What were you using to solve this before? (Open)
Think back to the last time you needed to do this. Walk us through what you actually did. (Open)
Pricing and value
At what price would this be so expensive that you would not consider it? (Numeric)
At what price would you start to question the quality? (Numeric)
Compared with what you pay today, this is: (Much cheaper / Slightly cheaper / About the same / Slightly more expensive / Much more expensive)
What would have to be true for the higher tier to be worth it to you? (Open)
Brand and market
When you think of [category], which brands come to mind? (Open, unaided, always before any brand list)
Which of these brands have you heard of? (Multi-select, randomized)
Which of these did you consider before choosing? (Multi-select)
In your own words, what does [brand] stand for? (Open)
Churn and win/loss
What is the main reason you are cancelling? (Single select, with "other, please specify")
Was there a specific moment when you decided to leave? What happened? (Open)
What are you switching to, if anything? (Open)
What would have kept you? (Open)
Screeners
Which of the following best describes your role in purchase decisions for [category]? (Single select: I decide / I influence the decision / I am not involved)
In the last 12 months, have you personally bought any of the following? (Multi-select, with decoy categories so the qualifying answer is not obvious)
Two notes on this list. The decoys in the last screener matter: a screener that makes the qualifying answer obvious invites people to select it. And every open question here is a specific prompt rather than a generic one, because "any other comments?" reliably returns nothing.
Common Survey Question Mistakes
The grid that eats the survey
A 12-item matrix on a 5-point scale is 12 questions dressed as one. On a phone it is nearly unusable, and it invites straightlining, where respondents pick the same column all the way down. Split long grids, cap them at roughly five or six rows, and check for straightlining in the data rather than assuming it did not happen.
Scales that do not balance
Positive and negative options should be symmetric in number and in intensity. "Excellent, very good, good, fair, poor" looks like a 5-point scale and behaves like a 3-point one, because four of the five labels are positive. Symmetric labels, with a defined midpoint if you want one.
Answers that are secretly the same
"Too expensive" and "poor value for money" in one list split the same underlying vote across two options, and both look smaller than the real thing. Read your option lists asking whether two of them mean the same to a respondent in a hurry.
The generic open question
"Any other comments?" gets low completion and vague text. Anchor open questions to something specific: the score just given, the last purchase, the moment of the decision. Then follow up on the answer, because almost every first open answer stops one level above the actual cause. "Delivery was slow" is a symptom; the useful information is which order, what was promised, and what they did instead.
Follow-up is the layer a static form cannot provide, which is why so much open text arrives as four words. This is where AI probing earns its place: a model that reads the first answer and asks one targeted second question turns a fragment into an account, without a moderator on the call. Voice widens it further, because people say considerably more out loud than they will type. We covered that in voice surveys vs traditional surveys.
Skipping the pilot
Twenty pilot responses will find your ambiguous stem, your missing answer option and your broken routing. All three are unfixable once real sample is in field. Read every pilot verbatim: confused answers point straight at the confusing question.
A Quick Checklist Before You Launch
Does every question map to a decision and an owner?
Does every question contain exactly one idea?
Does every closed question have an escape hatch for people who do not fit?
Is there a reference period on every behavioral question?
Are unaided questions before aided ones?
Are long unordered lists randomized?
Have you removed anything your own systems already know?
Have you taken the survey on a phone, yourself, end to end?
Where to Go Next
Question types in depth: Likert scales, matrix questions, rank order questions
Open versus closed: open ended questions and closed ended questions
Ready-made wording for CX: customer satisfaction survey questions
Design the whole study: data collection methods
Write fewer questions, learn more
Most survey drafts are long because the author is trying to anticipate every answer in advance. Twenty closed questions get written to cover the ground that two good open questions and a follow-up would have covered better.
SmartInterview lets you keep the survey short and still get the depth. Respondents answer by voice or text, an AI asks the follow-up a moderator would have asked, in their own language, and the open responses come back coded into themes you can count and trace to the exact words used.
Build your next questionnaire around the questions that matter and let the probing handle the rest. Start free or book a demo.
Frequently Asked Questions
What are the main types of survey questions?
Closed types are single select, multiple choice, dichotomous yes/no, Likert and rating scales, rank order and matrix grids. The open type lets respondents answer in their own words. Closed questions produce countable data; open questions produce explanations. Most good surveys are mostly closed with a small number of well-targeted open questions.
What makes a good survey question?
One idea, neutral wording, plain language, a stated reference period, and answer options that are exhaustive and mutually exclusive with an escape hatch for people who do not fit. Above all it has to map to a decision: if you cannot name who acts on the answer, the question does not belong in the survey.
How many questions should a survey have?
Length is better measured in minutes than in question count, since ten open questions take longer than thirty single-selects. A customer feedback survey typically sits at 3-6 questions. A market research study runs longer, but past roughly ten minutes expect data quality to fall as respondents speed up and stop reading carefully.
What are open survey questions and when should I use them?
Open questions let the respondent write or speak freely. Use them when you do not yet know the answer options, when you need the customer's own vocabulary, and when you need the reason behind a score. Keep them specific and anchored to something concrete, because generic prompts like "any other comments?" return generic text.
What is a leading question, and how do I avoid one?
A leading question signals the answer you want. It usually shows up as an adjective in the stem ("our improved checkout"), an unbalanced scale with more positive than negative options, or agree/disagree framing that invites agreement. Strip evaluative language from the stem and balance the scale on both sides.
What order should survey questions go in?
Screeners first, then an easy opener, then broad questions before narrow ones, the core measures while attention is high, sensitive questions late, and demographics last. Unaided awareness must always come before you show any brand or option list, because once respondents see the list you cannot unshow it.
How do I stop people from rushing through my survey?
Make it shorter, break long grids into smaller blocks, and ask about things the respondent actually cares about. Add attention checks sparingly, and look for straightlining and implausibly fast completions in the data rather than assuming the sample is clean.
