How to Write Survey Questions That Get Honest Answers

Writing survey questions that get honest answers comes down to a handful of rules: ask about one thing at a time, keep the wording neutral, balance your answer scales, and make the survey short enough that people finish it. Most bad survey data is not caused by dishonest respondents — it is caused by questions that accidentally push people toward an answer or make the honest answer hard to give.

Ask about one thing per question

A question that bundles two ideas — called a double-barreled question — forces respondents to average their feelings, and you cannot un-average the result.

• Bad: "How satisfied were you with the speed and quality of our support?"
• Good: "How satisfied were you with the speed of our support?" and, separately, "How satisfied were you with the quality of the answer you received?"

If someone got a fast but wrong answer, the bundled version gives you a meaningless 3. Two questions give you the actual story. When two questions feel like too many, cut the one you are less likely to act on rather than merging them.

Avoid leading and loaded wording

A leading question signals the answer you want; a loaded question smuggles in an assumption. Both inflate your scores and both are easy to write by accident.

• Bad (leading): "How much did you enjoy our redesigned dashboard?"
• Good: "How would you rate the redesigned dashboard?"

• Bad (loaded): "What did you find most confusing about setup?" (assumes setup was confusing)
• Good: "How easy or difficult was setup?" followed by an optional "What, if anything, was unclear?"

A quick self-test: could someone comfortably give a negative answer to this question? If the phrasing makes criticism feel rude, respondents will round up, and your data will read better than reality.

Use balanced scales

An unbalanced scale offers more positive options than negative ones, which tilts results before anyone answers.

• Bad: Excellent / Great / Good / Fair
• Good: Very satisfied / Satisfied / Neutral / Dissatisfied / Very dissatisfied

Balanced scales have a symmetric number of positive and negative options, usually around a neutral midpoint. Label the endpoints clearly ("1 = very difficult, 7 = very easy") so respondents do not have to guess which direction is good. Standard formats like the 1-5 satisfaction scale in CSAT or the 0-10 scale in NPS have this balance built in, which is a strong argument for using them instead of inventing your own.

Keep it short

Every added question costs you completions, and the people who drop out are not random — they are the busy, mildly-interested majority, leaving you with responses skewed toward the highly opinionated. Practical guidelines:

• Cut any question you do not have a plan to act on. "Interesting to know" is not a plan.
• Make open-text questions optional. Required essays are where completions go to die.
• Prefer one score plus one comment over five scores.
• If different people genuinely need different questions, branch instead of stacking. A chained survey with branching logic shows each respondent only the follow-up that fits their first answer, which keeps every individual path short.

Shorter surveys are also the single biggest lever on participation, which is covered in depth in how to increase survey response rates.

Put sensitive questions late

Questions about salary, demographics, criticism of a manager, or anything personally revealing belong at the end of the survey, for two reasons. First, respondents who have already invested a few minutes are more willing to answer them. Second, if someone does abandon at a sensitive question, you still keep everything they answered before it. Opening with "What is your age and income?" is the fastest way to lose someone who would happily have rated your product.

For genuinely sensitive topics — employee feedback especially — question wording alone is not enough; people need to believe the response cannot be traced to them. That is a collection-setup decision, covered in how to make an anonymous survey.

Match the question format to the job

The format shapes the answer as much as the wording does. A few pairings that consistently work:

• Quick reactions: emoji or star ratings — near-zero effort, good for in-the-moment pulses.
• Comparable measurement over time: standard scales (NPS, CSAT, 1-7 ease).
• Prioritization: ranking or multiple choice, not a stack of separate ratings.
• The "why" behind a score: one optional open-text follow-up.

Conejo Survey offers 30+ survey types, so the constraint is rarely the tool — it is choosing the lightest format that still answers your question.

Pilot before you send

Finally, test the survey on two or three people who were not involved in writing it, and ask them to think out loud while answering. You are listening for hesitation ("does 'support experience' mean the chat or the docs?"), for answers that do not fit the options offered, and for the moment their attention drops. Five minutes of piloting catches the ambiguous wording that would otherwise quietly corrupt hundreds of responses — and unlike most data problems, this one cannot be fixed after launch, because changing a live question splits your results into before and after.

Before sending any survey, read each question and ask: could a respondent comfortably give me bad news here? If not, rewrite it.

Where to go next

• How to increase survey response rates — good questions only matter if people answer them.
• How to make an anonymous survey — when honesty depends on the setup, not the wording.
• NPS vs CSAT vs CES — pick a standard metric before writing custom questions.