How to Write Survey Questions That Actually Get Responses
TL;DR
How to write survey questions that people actually answer comes down to a few principles: keep them short, avoid bias, and ask one thing at a time. Surveys with 1-3 questions achieve 83% completion rates, while surveys with 15+ questions drop to 42%.
Most survey questions fail not because the topic is wrong, but because the wording is bad.
Here's the quick version:
- Ask one thing per question -- double-barreled questions confuse respondents and produce unusable data
- Use neutral language -- leading questions inflate agreement rates by 10-15%
- Keep it under 5 minutes -- there's a 17% drop in response rate for surveys longer than 5 minutes
- Always include a follow-up "Why?" -- 44% of respondents will leave a comment if given the chance
- Design for mobile first -- nearly 60% of surveys are now completed on mobile devices
- Sometimes the best survey is no survey -- an open text box asking "What's on your mind?" often gets you more than 20 scaled questions
This guide walks through every common mistake, gives you side-by-side good vs. bad examples, and includes ready-to-use templates for NPS, CSAT, and CES surveys. If you're struggling with low response rates more broadly, check out our guide on how to get user feedback when nobody responds.
Why Do Most Feedback Survey Questions Fail?
Most feedback survey questions fail because they're easy to create but hard to do well. Anyone can type a question into a survey tool and hit send. Writing a question that produces useful, unbiased data is a different skill entirely.
As Erika Hall, co-founder of Mule Design and author of Just Enough Research, puts it: "Surveys are the one thing that really takes a lot of expertise to conduct and interpret correctly, and the ease of running a survey is completely out of proportion to the expertise required to actually do it right."
She's not wrong. Hall calls surveys "the most misused and misunderstood research tool" -- and the data backs her up.
Here's what goes wrong:
The Wording Problem
Pew Research Center has demonstrated repeatedly that small wording changes produce massive shifts in responses. When people were asked about government spending on "welfare," 44% said spending was too high. When the same question used "assistance to the poor," only 23% said the same thing. Same concept, different words, 21-point gap.
Another example: when asked about military action in Iraq, 68% favored it. When the question added "even if it meant that U.S. forces might suffer thousands of casualties," support dropped to 43%. The question changed -- the underlying opinion didn't.
The Length Problem
Survicate's research shows completion rates by survey length:
| Number of Questions | Completion Rate |
|---|---|
| 1-3 questions | 83.34% |
| 4-8 questions | 65.15% |
| 9-14 questions | 56.28% |
| 15+ questions | 41.94% |
Every question you add costs you completions. The relationship isn't linear either -- the biggest drop happens in the first few additional questions. Once someone commits to answering 15 questions, adding more up to 35 causes smaller incremental drops, but getting them past that first threshold is the hard part.
The Bias Problem
The worst part about biased survey questions? People will still answer them. Hall points out that "even if you're asking the wrong questions in the wrong way, and they totally don't match anything in the real world, you'll get answers back." Unlike an interview where awkward silence tells you something's off, a biased survey returns confident-looking data that's actually garbage.
Understanding the difference between qualitative and quantitative feedback helps you decide when a survey is even the right tool for what you're trying to learn.
What's the Difference Between Open and Closed Survey Questions?
Open-ended questions let respondents answer in their own words. Closed-ended questions give them predefined choices. Both have a role, but using the wrong type at the wrong time wastes everyone's effort.
Here's how they compare:
| Criteria | Open-Ended Questions | Closed-Ended Questions |
|---|---|---|
| Response format | Free text | Predefined options (scales, multiple choice, yes/no) |
| Data type | Qualitative | Quantitative |
| Analysis effort | High (manual coding or AI) | Low (automatic aggregation) |
| Completion speed | Slower | Faster |
| Insight depth | Deep -- uncovers surprises | Surface -- confirms hypotheses |
| Response rate impact | Lowers completion if overused | Higher completion rates |
| Best for | Understanding "why" | Measuring "how much" or "how many" |
When to Use Open-Ended Questions
Use open-ended questions when you don't know what you don't know. As Nielsen Norman Group notes, the greatest benefit of open-ended questions is that they let you discover things you didn't anticipate. If you're exploring a problem space, closed questions limit you to the answers you already imagined.
Good open-ended questions:
- "What's the biggest challenge you face with [problem area]?"
- "What almost stopped you from completing this task?"
- "If you could change one thing about [product/feature], what would it be?"
When to Use Closed-Ended Questions
Use closed-ended questions when you need measurable, comparable data. They're faster to answer, easier to analyze, and better for tracking trends over time. SurveyMonkey recommends closed-ended questions for most quantitative research because they produce consistent, statistically analyzable data.
Good closed-ended questions:
- "How satisfied were you with this experience?" (1-5 scale)
- "Which feature do you use most often?" (multiple choice)
- "Would you recommend us to a colleague?" (0-10 NPS scale)
The Best Approach: Combine Both
The most effective surveys pair a closed-ended question with an open-ended follow-up. Ask "How satisfied were you?" (closed), then "What's the main reason for your rating?" (open). You get the measurable score and the context behind it.
How Do You Avoid Leading and Biased Survey Questions?
Leading questions push respondents toward a specific answer, inflating or deflating your results without you realizing it. Acquiescence bias alone inflates agreement rates by 10-15% on typical agree/disagree scales, and that's before you add loaded language.
Here are the four most common bias traps -- and how to fix each one.
1. Leading Questions (Loaded Language)
Leading questions use adjectives or framing that suggest the "right" answer.
| Bad (Leading) | Good (Neutral) |
|---|---|
| "How great was your experience with our support team?" | "How would you rate your experience with our support team?" |
| "How much do you love our new dashboard?" | "How would you describe your experience with the new dashboard?" |
| "Don't you think we should add dark mode?" | "How important is a dark mode option to you?" |
| "How helpful was our amazing onboarding guide?" | "How would you rate the onboarding guide?" |
The fix is simple: remove subjective adjectives. Words like "great," "amazing," "exciting," "innovative" -- they all push respondents toward agreement.
2. Double-Barreled Questions (Two Questions in One)
Double-barreled questions ask about two things at once, making the response impossible to interpret. Qualtrics warns that these produce ambiguous data because you can't tell which part the respondent is reacting to.
| Bad (Double-Barreled) | Good (Split Into Two) |
|---|---|
| "How satisfied are you with our product quality and customer service?" | "How satisfied are you with our product quality?" + "How satisfied are you with our customer service?" |
| "Was the agent friendly and efficient?" | "How would you rate the agent's friendliness?" + "How would you rate the agent's efficiency?" |
| "Is our pricing fair and transparent?" | "How fair do you find our pricing?" + "How clear is our pricing information?" |
If a question contains "and" or "or" connecting two topics, it's probably double-barreled. Split it.
3. Acquiescence Bias (The "Agree" Trap)
People have a natural tendency to agree with statements, especially in surveys. Pew Research Center found that when respondents were asked if they agree that "the best way to ensure peace is through military strength," 55% agreed. But when given a forced choice between that statement and "diplomacy is the best way to ensure peace," only 33% chose military strength while 55% chose diplomacy.
The fix: instead of "agree/disagree" formats, offer balanced alternatives. Instead of "Do you agree that our onboarding is easy?", ask "How would you rate the difficulty of our onboarding process?" on a scale from "Very difficult" to "Very easy."
4. Hypothetical Questions (Predicting Future Behavior)
Rob Fitzpatrick, author of The Mom Test, built an entire framework around this problem: people are terrible at predicting their own future behavior. Asking "Would you use this feature?" produces unreliable data because respondents imagine an ideal version of themselves.
| Bad (Hypothetical) | Good (Past Behavior) |
|---|---|
| "Would you pay for a premium plan?" | "Have you purchased a premium plan for any software tool in the past year?" |
| "Would you use an AI categorization feature?" | "How do you currently categorize your feedback?" |
| "How likely are you to recommend us?" | This one's fine -- it's the established NPS question with decades of benchmarking data behind it |
The fix: ask about past behavior, not future intentions. What someone has done is a much better predictor than what they say they'll do.
How Does Survey Length Affect Completion Rates?
Shorter surveys get more responses. The data on this is unambiguous. SurveyMonkey found a 17% drop in response rate when surveys exceed 12 questions or take longer than 5 minutes. Survey abandon rates increase sharply for surveys taking more than 7-8 minutes.
Here's a practical guide for different survey types:
| Survey Type | Optimal Length | Time Target | Expected Completion Rate |
|---|---|---|---|
| NPS survey | 1 question + 1 follow-up | 30 seconds | 85-90% |
| CSAT (post-interaction) | 1-3 questions | 1-2 minutes | 75-83% |
| CES (effort score) | 1-2 questions | 30-60 seconds | 80-85% |
| Feature feedback | 3-5 questions | 2-3 minutes | 65-75% |
| Onboarding feedback | 3-5 questions | 2-3 minutes | 60-70% |
| Product research | 8-12 questions | 5-7 minutes | 50-60% |
| Market research | 10-15 questions max | 7-10 minutes | 40-50% |
The guideline from HubSpot's research is the "10/10 rule": no more than 10 questions, taking no more than 10 minutes. For transactional feedback (post-purchase, post-support), aim for 1-4 questions in about 2 minutes.
If you need a longer survey, be upfront about the time commitment. Saying "This will take about 5 minutes" actually increases completion because respondents can make an informed choice about whether to start.
Not sure which metric to anchor your survey around? Our NPS vs CSAT vs CES comparison breaks down when each makes sense.
What Are the Best Survey Question Templates for NPS, CSAT, and CES?
The three standard feedback survey question frameworks -- NPS, CSAT, and CES -- each serve different purposes. Here are ready-to-use templates you can deploy today, with the follow-up questions that actually produce actionable insights.
NPS (Net Promoter Score) Template
Primary question:
"On a scale of 0-10, how likely are you to recommend [Product Name] to a friend or colleague?"
Follow-up for Promoters (9-10):
"What do you love most about [Product Name]?"
Follow-up for Passives (7-8):
"What would make you rate us higher?"
Follow-up for Detractors (0-6):
"What's the main reason for your score? We'd genuinely like to understand."
CSAT (Customer Satisfaction) Template
Primary question:
"How satisfied were you with [specific interaction/feature/experience]?" Scale: Very Unsatisfied (1) - Very Satisfied (5)
Follow-up:
"What could we have done better?"
CES (Customer Effort Score) Template
Primary question:
"To what extent do you agree: [Product Name] made it easy for me to [specific task]." Scale: Strongly Disagree (1) - Strongly Agree (7)
Follow-up:
"What made this difficult? (if anything)"
Which Scale Should You Use: 5-Point or 7-Point?
Research from MeasuringU shows 7-point scales are slightly more accurate than 5-point scales. On 5-point scales, 2.5% of responses involved users interpolating between points (writing "3.5" when they wanted something between 3 and 4). The 7-point version eliminated this entirely.
However, 5-point scales have higher completion rates and are easier for respondents to process. The practical advice: use 5-point scales for CSAT (simplicity wins) and 7-point scales for CES (you need the granularity). NPS always uses 0-10.
How Does Question Order Affect Survey Results?
Question order influences responses more than most people realize. Placing a specific question before a general one changes how people answer the general question, because the specific question primes their thinking.
Research compiled by SurveyMonkey shows two key effects:
Primacy bias -- in visual surveys (web, mobile), respondents tend to pick options listed first. When reading a list, people give more weight to early items because they evaluate each option before seeing the rest.
Recency bias -- in verbal surveys (phone, in-person), respondents favor the last option mentioned. They remember the most recent thing they heard.
The Funnel Technique
The most reliable approach is the funnel technique, recommended by Qualtrics:
- Start broad -- general satisfaction or experience questions
- Narrow down -- specific feature or interaction questions
- Get personal -- demographics and open-ended questions last
This works because broad questions aren't contaminated by specific priming. If you ask "How do you feel about our billing page?" first, their answer to "How satisfied are you with our product overall?" will be disproportionately influenced by billing.
Where to Place Demographics
Pew Research Center and most survey methodology experts recommend saving demographic questions (age, role, company size) for the end. These questions feel invasive early on and can trigger privacy concerns that increase abandonment.
The exception: if you need demographics for skip logic (routing different segments to different questions), you may need one or two screening questions upfront.
Randomize When Possible
For multiple-choice questions with several options, randomize the order. This neutralizes primacy bias. Most survey tools (including alternatives to Typeform) support answer randomization natively.
How Do You Write Survey Questions for Mobile Users?
Design for mobile first, because that's where most people will take your survey. SurveyMonkey's State of Surveys report found that nearly six out of ten surveys were completed on mobile devices in 2024, and that percentage keeps climbing. SurveySparrow reports that conversational mobile surveys can reach 85% completion rates -- but only if they're designed for the format.
Mobile Survey Design Rules
Keep questions short. On a small screen, a question that wraps to 4-5 lines feels overwhelming. Aim for questions that fit in 1-2 lines on a mobile screen.
Use tappable answer options. Radio buttons and small checkboxes are frustrating on touchscreens. Use large, tappable buttons or card-style answer options.
Limit matrix questions. Those grid-style "rate each item on a scale" tables are nearly impossible on mobile. Each row requires horizontal scrolling. Either break them into individual questions or use a different format entirely.
One question per screen. Scrolling through a long page of questions feels tedious on mobile. Showing one question at a time with a progress indicator creates a sense of momentum.
Minimize typing. Open-ended questions on mobile get shorter, less detailed responses because typing on a phone is slower. If you need qualitative feedback on mobile, consider voice input options or keep the text field optional.
Surveys longer than 9 minutes on mobile see high drop-off rates. For mobile-first audiences, keep the entire experience under 3 minutes. If you're still figuring out where and how to trigger these surveys in the first place, our walkthrough on how to set up user feedback collection in 5 minutes covers the basics. Choosing the tool that will show them in-app? Our comparison of Survicate alternatives covers the options for in-app feedback.
Why Are Follow-Up Questions the Most Powerful Part of Any Survey?
The single most valuable question in any survey isn't the rating -- it's the follow-up "Why?" A score tells you that someone is happy or unhappy. The follow-up tells you why, which is what you actually need to improve.
CustomerGauge reports that 44% of survey respondents will leave a comment when given an open-ended follow-up option. That's nearly half your respondents volunteering qualitative context for free.
How to Write Effective Follow-Up Questions
The best follow-up questions are specific to the score, not generic.
Generic (less effective):
"Any additional comments?"
Score-specific (more effective):
- For high scores: "What's the main thing you value about [product]?"
- For low scores: "What's the one thing we could improve?"
- For mid-range scores: "What would need to change for you to rate us higher?"
The "One Thing" Technique
Notice the questions above ask for "the one thing" or "the main thing." This constraint is intentional. Asking "What could we improve?" invites a rambling list. Asking for "the one thing" forces respondents to prioritize, giving you their highest-impact insight.
When to Make Follow-Ups Required vs. Optional
Make follow-ups optional. Required open-ended fields on surveys lead to junk responses ("asdf," "nothing," "fine") from people who just want to finish. An optional field with a well-worded prompt gets fewer but dramatically better responses.
The insights from follow-up questions are where the real product decisions come from. If you're looking for how to turn those open-ended responses into a clear product direction, our feedback chaos to product clarity playbook covers the full process.
Good vs. Bad Survey Questions: Side-by-Side Examples
Here's a reference table of common survey question mistakes with their fixes. Bookmark this -- you'll come back to it.
| Problem Type | Bad Question | Why It's Bad | Better Question |
|---|---|---|---|
| Leading | "How much do you enjoy our product?" | Assumes enjoyment | "How would you describe your experience with our product?" |
| Double-barreled | "Is our app fast and easy to use?" | Asks two things | "How would you rate the speed of our app?" (separate question for ease of use) |
| Loaded language | "Do you agree that our revolutionary AI saves you time?" | "Revolutionary" biases toward agreement | "Has the AI feature changed the time you spend on [task]? If so, how?" |
| Hypothetical | "Would you use a Slack integration?" | People can't predict future behavior | "How do you currently share feedback with your team?" |
| Vague | "How do you feel about our service?" | Too broad to be actionable | "How satisfied were you with the response time on your last support ticket?" |
| Absolute | "Do you always check your analytics dashboard?" | "Always" forces false precision | "How often do you check your analytics dashboard?" (with frequency options) |
| Jargon-heavy | "How would you rate the UX of our PLG onboarding funnel?" | Not everyone knows your acronyms | "How easy was it to get started with [Product Name]?" |
| Negative framing | "What problems have you had with our product?" | Assumes problems exist | "How has your experience been with our product so far?" |
The Readability Rule
Pew Research Center recommends that survey language should be accessible to all respondents. Industry best practice is to write at a 6th-grade reading level, using short words and simple sentence structures. If your survey question contains words like "utilize," "leverage," or "facilitate," replace them with "use," "help," or "make easier."
Putting It All Together: Your Survey Question Checklist
Before you send any survey, run through this list:
Question Quality:
- Each question asks only one thing (no double-barrels)
- No leading language or subjective adjectives
- Written at a 6th-grade reading level
- Questions about past behavior, not hypothetical futures
- Answer options are balanced (not skewed toward positive)
Survey Structure:
- 5 questions or fewer for transactional surveys
- General questions before specific ones (funnel technique)
- Demographics at the end
- Answer options randomized where appropriate
- Progress indicator visible
Follow-Up:
- At least one open-ended follow-up after the main rating
- Follow-up is optional, not required
- Follow-up prompt is specific ("What's the one thing...") not generic ("Any comments?")
Mobile:
- Questions fit in 1-2 lines on mobile
- Tappable answer buttons (not tiny radio buttons)
- No matrix/grid questions
- Total time under 3 minutes on mobile
After the Survey:
- Plan to close the feedback loop with respondents
- System to analyze open-ended responses
- Process to act on what you learn
Collect Better Feedback with FeedSense
Here's an honest take: sometimes the best survey is no survey at all. A simple open text box asking "What's on your mind?" -- shown at the right moment, in the right context -- can outperform a 15-question survey any day.
That's the approach FeedSense takes. Instead of bombarding users with long surveys, we help you:
- Ask simple, focused questions at contextual moments inside your product
- Collect open-ended feedback and use AI to categorize and analyze it automatically
- Track every response so you can close the loop with the people who took time to answer
- See sentiment trends without requiring users to fill out scaled questionnaires
We're not anti-survey. NPS, CSAT, and CES all have their place. But we believe the richest feedback comes from making it easy for users to tell you what they think in their own words -- and then actually doing something with it.
Try FeedSense free -- free tier includes 100 feedback items/month, no survey fatigue required.
FAQ
How many questions should a feedback survey have?
For transactional feedback (post-support, post-purchase), keep it to 1-3 questions. For product research, stay under 10. Studies show surveys with 1-3 questions achieve 83% completion rates, while 15+ question surveys drop to 42%. The sweet spot for most SaaS feedback is 2-5 questions that take under 3 minutes.
What is the difference between open-ended and closed-ended survey questions?
Open-ended questions let respondents answer in their own words ("What could we improve?"), producing qualitative insights. Closed-ended questions offer predefined choices ("Rate 1-5"), producing quantitative data. Use closed-ended for measurement and benchmarking, open-ended for discovery and understanding the "why" behind scores.
How do you avoid leading questions in a survey?
Remove subjective adjectives ("great," "amazing," "innovative") from your questions. Use neutral wording like "How would you rate..." instead of "How much did you enjoy..." Offer balanced answer scales, and avoid agree/disagree formats, which inflate agreement by 10-15% due to acquiescence bias.
Does survey question order matter?
Yes. Specific questions asked before general ones prime respondents' thinking, skewing their answers. Use the funnel technique: start with broad questions, then narrow to specifics, and save demographics for last. Randomize answer options to prevent primacy bias, where people tend to pick the first option listed.
What are the best survey question templates for customer feedback?
NPS uses "How likely are you to recommend us?" (0-10). CSAT uses "How satisfied were you with [specific experience]?" (1-5). CES uses "To what extent do you agree: [Company] made it easy to [task]" (1-7). Always pair these with an open-ended follow-up like "What's the main reason for your score?"
How long should a survey take to complete?
Under 5 minutes for most surveys. Research shows a 17% drop in response rates when surveys exceed 5 minutes or 12 questions. On mobile, aim for under 3 minutes. For quick pulse checks like NPS, 30 seconds is ideal. If your survey must be longer, tell respondents upfront how long it will take.
Should I use a 5-point or 7-point scale for survey questions?
Five-point scales are simpler and have higher completion rates. Seven-point scales are slightly more accurate and eliminate the problem of respondents trying to mark between points. Use 5-point for CSAT (simplicity wins), 7-point for CES (you need the granularity), and 0-10 for NPS (it's the standard).
What is a double-barreled survey question?
A double-barreled question asks about two things at once, like "Is our app fast and easy to use?" If someone finds it fast but hard to use, they can't answer accurately. The fix: split it into separate questions. Any question with "and" or "or" connecting two concepts is likely double-barreled.
How do I write survey questions for mobile users?
Keep questions to 1-2 lines on a mobile screen. Use large tappable buttons instead of small radio buttons. Show one question per screen with a progress indicator. Avoid grid/matrix questions, which require horizontal scrolling. Minimize typing by making open-ended responses optional.
Why do people abandon surveys before finishing?
The top reasons are length (surveys over 7-8 minutes see sharp drop-offs), irrelevant questions (not using skip logic), bad mobile design (tiny buttons, grid questions), and no clear purpose. If respondents don't understand why you're asking or how long it will take, they'll quit.
Sources
- Pew Research Center -- Writing Survey Questions
- Pew Research Center -- Survey Experiments and Question Wording
- Survicate -- Survey Completion Rate Study
- SurveyMonkey -- Survey Completion Times and Abandon Rates
- SurveyMonkey -- State of Surveys 2025
- SurveyMonkey -- Eliminate Order Bias
- SurveyMonkey -- Open vs Closed-Ended Questions
- SurveySparrow -- Mobile Survey Completion Rates
- SurveySparrow -- Biased Survey Questions
- Qualtrics -- Acquiescence Bias
- Qualtrics -- Double-Barreled Questions
- Qualtrics -- Question Sequence, Flow, and Style
- CustomerGauge -- NPS Survey Best Practices
- MeasuringU -- 5 or 7 Point Scales
- Nielsen Norman Group -- Survey Best Practices
- Nielsen Norman Group -- Open-Ended Questions
- HubSpot -- Ideal Survey Length
- Erika Hall -- Just Enough Research / Mule Design
- Erika Hall on User Interviews Podcast
- Rob Fitzpatrick -- The Mom Test
Tags:
Stay in the loop
Get the latest insights on customer feedback, product management, and building better SaaS products. No spam, unsubscribe anytime.
By subscribing, you agree to our privacy policy. We respect your inbox.