Guide
Designing a Customer Survey That Gets Honest Answers
A survey is only as good as the answers it collects. This guide shows you how to write questions that people can answer quickly and truthfully, so the results actually help you make a decision.
Start with the decision, not the questions
Before you write a single question, write one sentence that begins: “After this survey, I will decide…” For example: “After this survey, I will decide whether to add a weekday lunch combo.” That sentence becomes your filter. Every question must help with that decision. If it doesn’t, cut it.
Why people give answers that aren’t quite true
Most people aren’t lying on purpose; they’re being polite, rushing or guessing. Know the usual culprits:
- Politeness (social desirability). People give the answer that sounds good. Ask “Would you buy organic vegetables?” and many will say yes because it sounds responsible.
- Agreeing by default (acquiescence). Long lists of “Do you agree that…” statements push people to tick “agree” without thinking.
- Leading wording. “How much did you enjoy our new menu?” assumes they enjoyed it.
- Fuzzy memory. “How many times did you eat out last year?” is close to impossible to answer accurately.
- Imagining the future. People are poor at predicting their own behaviour, especially their spending.
Writing questions that work
Ask about past behaviour, not future intentions
“Would you pay for home delivery?” invites a hopeful guess. “In the last two weeks, how many times did you order groceries for delivery?” asks for a fact the person actually knows. Recent, specific behaviour is the most reliable thing a survey can measure.
One idea per question
“Was the staff friendly and quick?” is really two questions. Someone who found them friendly but slow can’t answer honestly. Split it.
Use neutral wording and balanced scales
Offer as many negative options as positive ones, with a clear middle: Very dissatisfied, Dissatisfied, Neither, Satisfied, Very satisfied. Avoid scales like “Good / Very good / Excellent”, which leave no way to say “bad”.
Give people a way out
Add “Don’t know”, “Not applicable” or “Prefer not to say” where they make sense. Forced answers just add noise.
Make anonymity real
If you don’t need names or phone numbers, don’t ask for them. People are more candid when they can see they cannot be identified.
Before and after: fixing common questions
| Weak question | Problem | Better version |
|---|---|---|
| “Don’t you think our prices are reasonable?” | Leading; invites agreement | “Compared with similar shops nearby, are our prices…” (Much lower to Much higher) |
| “How often do you shop online?” | Vague time frame and terms | “In the past 30 days, how many separate orders did you place on any shopping app or website?” |
| “Would you use an app to book our salon?” | Asks people to predict | “How did you book your last salon visit?” (Walked in / Phone call / WhatsApp / App / Other) |
| “Rate our quality and service.” | Two ideas in one | Two separate questions: one for quality, one for service |
| “What is your income?” (open box) | Sensitive and hard to answer | Broad bands, plus “Prefer not to say” |
Question order and length
Open with easy, non-threatening questions that relate to the topic. Put sensitive questions, such as income, near the end, and only if you truly need them. Ask general questions before specific ones, so that a detailed question about your brand doesn’t colour answers to a broader question about the category. Aim for about five minutes on a phone, and time yourself.
Language and channels in India
Write in the language your customers are most comfortable reading, and offer more than one if your customers are mixed. When translating, ask a second person to translate the result back into the original language; any differences show you where the meaning slipped. Most people will open your link on a phone, often from WhatsApp, so check how every question looks on a small screen. For customers less comfortable online, a short paper form at the counter can reach people a link never will.
Pilot test before you launch
- Ask five to eight people who resemble your real respondents to take the survey while you watch or listen.
- Ask them to think aloud: “What do you think this question is asking?”
- Note any question where they hesitate, re-read, or understand something different from what you meant.
- Rewrite those questions, and cut any that don’t serve your decision sentence.
- Check the timing, and test the link on at least two different phones.
Pre-launch checklist
- I can state the decision this survey informs in one sentence.
- Questions ask about specific, recent behaviour where possible.
- No question contains two ideas or leading wording.
- Scales are balanced and include a neutral or “Don’t know” option where sensible.
- Sensitive questions are optional and placed at the end.
- The survey takes about five minutes on a phone.
- I have piloted it and fixed the confusing questions.
- I tell respondents why I’m asking and how their answers will be used.