How to write good survey questions
By Thisys team · Last reviewed
Most bad survey data starts with a question that could be read two ways, nudged people towards an answer, or offered no answer that fitted. These rules fix the common faults, and each comes with a weak example and a better one.
Start from the decision
Before you write a question, list the decisions the survey will inform. Every question should feed one of them. If you would not do anything differently whatever the answer, cut the question: a shorter survey gets more complete answers.
Ask one thing at a time
| Weak | Better |
|---|---|
| How satisfied are you with our prices and service? | How satisfied are you with our prices? (then a separate question about service) |
Someone who likes the service and not the prices cannot answer the weak version honestly, and you cannot tell which half their answer is about.
Do not lead
| Weak | Better |
|---|---|
| How much did you enjoy our new menu? | How would you rate our new menu? |
| Most people think remote work is good for productivity. Do you agree? | What effect, if any, has remote work had on your productivity? |
The weak versions assume the answer. People tend to go along with a question's framing, so the result measures the wording as much as the opinion.
Avoid loaded words
| Weak | Better |
|---|---|
| Do you support the council wasting money on a new stadium? | Do you support or oppose the council building a new stadium? |
Name both sides of a choice (support or oppose, agree or disagree) and leave out words that carry a verdict.
Give answers that fit everyone
| Weak | Better |
|---|---|
| 0–5, 5–10, 10–20 | Under 5, 5 to 9, 10 to 19, 20 or more |
Answer options should not overlap and should cover everyone. Where some people will not fit, add “Not applicable”, “Other (please specify)” or “None of these” rather than forcing a wrong answer.
Balance your scales
| Weak | Better |
|---|---|
| Excellent, Very good, Good, Fair | Very poor, Poor, Fair, Good, Very good |
The weak scale has three positive points and nothing below Fair, so it pushes ratings up. Give equal room to both sides, label at least the two ends, and keep the direction the same throughout the survey.
Use plain words and a clear time frame
| Weak | Better |
|---|---|
| How often do you use our app? | In the past 30 days, on about how many days did you use our app? |
| Do you disagree that staff should not be allowed to work from home? | Should staff be allowed to work from home? |
“Often” means different things to different people; a time frame and a count do not. Spell out abbreviations, avoid jargon, and never make people untangle a double negative.
Do not rely on long memories
The further back you ask people to remember, the less reliable the answer. Ask about a recent, specific period (“in the past week”) rather than a typical year, and ask what people did rather than what they would do.
Put sensitive questions last
Open with easy, relevant questions. Put demographic and sensitive questions near the end, make them optional where you can, say why you are asking, and offer “Prefer not to say”.
Test it before you send it
- Answer the whole survey yourself on a phone, and time it.
- Ask five people from the audience to answer it while saying aloud what each question means to them.
- Fix every question someone hesitated over, misread or skipped.
- Check the test answers come out in the form you planned to analyse.
Use this template
Each opens its own page, where you can preview every question before you use it.
Customer Satisfaction Survey
Preview and use this template →
Employee Engagement Survey
Preview and use this template →
Market Research Survey
Preview and use this template →
Use cases: Product feedback surveys and Academic research questionnaires.
Read next
Put it into practice. Free to start, and nothing to install.