100+ survey question examples for customer satisfaction, NPS, product feedback, churn, events, employees and market research, plus the mistakes to avoid.
A survey is only as good as its questions. Ask something vague and you get vague answers. Ask two things at once and you get answers you cannot use. Ask a leading question and you get the answer you were hoping for, which is worse than no answer at all.
This page is a reference: more than 100 survey question examples, grouped by what you are trying to learn, with a note on which question type fits and when to send it. Copy them, adapt the wording to your product or audience, and cut anything you will not act on.
If you want to turn a list like this into a live survey quickly, Fast Surveys has an AI survey generator: describe what you want to learn and it drafts a structured survey you can then edit question by question. It is free up to 50 responses per survey.
The same question can work or fail depending on the format you pick, so here are the six you will use most.
The respondent picks one option or several from a list you define. Fast to answer, easy to analyze, and right when you already know the likely answers. Add an "Other" option with a text field whenever you are not sure the list is complete.
Example: "How did you first hear about us?" Search engine / Social media / A friend or colleague / Newsletter or blog / Other
A rating scale asks for a score (1 to 5 stars, 1 to 10). A Likert scale asks how much someone agrees with a statement, usually on five points from "Strongly disagree" to "Strongly agree". Both measure intensity and track change over time well, as long as the scale stays identical between surveys. Label at least both ends, so a "3" means the same thing to everyone.
Example: "The setup instructions were easy to follow." Strongly disagree / Disagree / Neutral / Agree / Strongly agree
Mood scales (a row of faces or emoji) are a lighter variant: less precise, but quick and friendly for short surveys shown in context.
One specific question, "How likely are you to recommend product to a friend or colleague?", answered on a 0 to 10 scale. It gets its own section below, because it is scored differently from a normal rating.
A text box. Open questions give you the words your customers actually use, the reasons behind a score, and the things you did not think to ask about. The cost is effort: they are slower to answer and slower to analyze. Use one or two per survey, ideally right after a closed question ("What is the main reason for your score?").
Binary questions are the easiest to answer and to count. Use them for facts ("Did you contact support in the last 30 days?") and to route people to the right follow-up questions. Avoid them for opinions, where the honest answer is rarely a clean yes or no.
The respondent orders a list from most to least important. Ranking forces trade-offs, which a rating scale does not: given the chance, people rate every feature "very important". Keep the list to five to seven items, or the middle turns into noise.
Example: "Rank these in order of importance to you: price, speed, integrations, customer support, ease of use."
Customer satisfaction (CSAT) questions measure how people feel about a specific interaction or about your product as a whole. Keep the survey short and send it close to the moment you are asking about.
Question 5 is a customer effort question. It is often more useful than satisfaction itself, because "this was hard" points straight at something you can fix.
NPS is a standard question with a standard scoring method, which is what makes it useful for comparing yourself over time. As Bain & Company defines it, people who answer 9 or 10 are promoters, 7 or 8 are passives, and 0 to 6 are detractors. Your score is the percentage of promoters minus the percentage of detractors, so it lands anywhere between -100 and +100. Once your responses are in, calculate your score with our free NPS calculator.
The score on its own tells you very little. The follow-up question is where the value is.
Question 18 is a common product/market fit check: the share of people who would be "very disappointed" shows how much your users actually rely on you, which a satisfaction score does not.
If you want NPS answers from people while they are on your site, rather than from whoever opens an email a week later, Fast Surveys offers NPS as one of its website feedback widget formats, and the answers land in the same dashboard as your surveys.
These go to existing users. The goal is to learn what to build, fix or remove, so ask about behavior and problems rather than hypothetical features.
Be careful with "Would you use a feature that does X?". Almost everyone says yes to a free hypothetical. Ask about the problem instead: "How do you handle X today?" What people already do is a much better signal than what they say they might do.
Website questions work best in context. A small prompt on the page, shown at the right moment, beats an email a week later when nobody remembers what the page looked like. Stick to one or two questions per prompt.
This is exactly what Fast Surveys' website feedback widgets are for. You paste a small snippet into your site, pick a format (quick choices, mood ratings or written feedback), and every response goes to the same place as your regular surveys, so "Was this article helpful?" on your docs and your quarterly customer survey sit side by side.
Onboarding questions do two jobs: they help you tailor the first experience, and they show you where new users get stuck. Ask the profile questions at sign-up (and keep them short), and the experience questions after a few days of use.
A cancellation survey is the last conversation you will have with a customer, so make it short and respectful. One required multiple choice question and one optional text field are enough. Never make people fill in a form before they are allowed to leave.
Question 56 tends to be the most revealing. If a lot of people leave for the same competitor, that tells you more than any satisfaction score.
Send it within a day, while people still remember the details, and split your questions between content, logistics and the event overall. For an in-person event, a QR code on the last slide or on the tables lets people answer before they leave instead of digging through their inbox later.
Employee surveys only work if people trust that their answers are anonymous and that something will happen afterwards. Say who will see the results, then share what you learned and what you are going to change. Most of these work best as Likert statements.
Market research surveys usually go to people who are not your customers yet. Screen first, so you know who answered, then ask about the problem and their current behavior before you ever mention your product.
If you run a newsletter, a YouTube channel or a podcast, an audience survey tells you what to make next and which sponsors your audience will actually welcome. Keep it under five minutes, and say what you will do with the answers.
If a full survey feels like too much, ask one of these per issue. You will build the same picture over a few weeks.
A leading question suggests the answer. "How much did you enjoy our new dashboard?" assumes the person enjoyed it. Ask "How would you rate the new dashboard?" instead.
A loaded question contains an assumption that may not be true. "What do you like most about our fast delivery?" is useless to someone whose order arrived late. Check the assumption first with a yes/no question, then ask the follow-up only to the people it applies to.
"How satisfied are you with our pricing and customer support?" is two questions. Someone who loves your support and hates your pricing has no honest answer. Split it in two.
"Always", "never", "every" and "all" push people towards no. "Do you always read our newsletter?" gets a no from someone who reads nine issues out of ten. Ask about frequency instead.
"Poor / Good / Very good / Excellent" has three positive options and one negative, so it inflates your results. Give the scale as many points on the negative side as on the positive side.
"0-10 employees" and "10-50 employees" overlap. A list without "Other" or "Not applicable" forces people to pick something false. Both give you data you cannot trust.
Your team knows what "the Insights tab" or "SSO" means. Many of your respondents may not, and they will answer anyway. Use the words your customers use.
"Would you pay $10 a month for this?" measures politeness. "What did you last pay for something like this?" measures behavior.
Every question should map to a decision you will make. If you would not act on the answer, cut it. Put your most important question first in case people drop off, say how long the survey takes, and make sure that is true.
Forcing someone to type an answer mostly gets you "n/a" and "asdf". Keep text fields optional, and they will be filled by the people who have something to say.
Once you have your questions, the process is short:
For the building part, Fast Surveys is a survey builder made in the UK that covers most of this page. It has multiple choice, checkboxes, dropdowns, mood scales, text fields and image questions, an AI generator that drafts a survey from a short description, and AI summaries of the written answers once they come in. You can share by link or QR code, embed feedback widgets on your website, match your brand colours, and invite unlimited team members. The free plan includes unlimited surveys with up to 50 responses per survey and a limited set of question types. The Pro plan ($29 a month at the time of writing) unlocks all question types and unlimited responses. Fast Surveys is a Uneed partner.
Want to compare options first? Browse the survey tools listed on Uneed, or, if you are moving away from Typeform, see our Typeform alternatives.
A good survey question is specific, neutral, and asks about one thing at a time. The respondent can answer it from their own experience, and the answer helps you make a decision. If a question fails any of those, rewrite it or cut it.
As few as you need to make your decision. A widget on a web page should ask one or two. A post-purchase or post-support survey works well with one to three. A yearly customer or employee survey can be longer, as long as you tell people upfront how long it will take.
Closed-ended questions (multiple choice, rating, yes/no, ranking) offer fixed options, so answers are easy to count and compare. Open-ended questions let people answer in their own words: slower to analyze, but they tell you why. Most good surveys are mostly closed, with one or two open questions.
A Likert scale question presents a statement and asks how much the respondent agrees with it, usually on five points: Strongly disagree, Disagree, Neutral, Agree, Strongly agree. It is the standard format for employee engagement and attitude surveys.
Take the percentage of respondents who answered 9 or 10 (promoters) and subtract the percentage who answered 0 to 6 (detractors). Passives (7 or 8) count in the total but in neither group. For example, if 200 people answer, 120 are promoters (60%) and 30 are detractors (15%), your NPS is 60 minus 15, so 45. Or skip the arithmetic with our free NPS calculator.
Keep it short and say how long it takes. Ask close to the experience you are asking about, and reach people where they already are. Then show that you acted on the answers: people who see their feedback lead to a change are more willing to answer next time.
