How to make a questionnaire that gets honest answers

A questionnaire is the instrument; a survey is the exercise of sending it to people. The distinction matters because it changes what you optimise for: a questionnaire has to be answerable consistently by everyone who reads it, which is a stricter requirement than merely being clear.

Start with your variables, not your questions

List the things you need to measure before writing anything anyone will read. Each variable then gets exactly one question, or one deliberate group of questions. Working the other way round — brainstorming questions and hoping they cover it — produces questionnaires that ask the same thing three times and miss something important entirely.

One idea per question

The most common flaw in an amateur questionnaire is the double-barrelled question. "Was the venue clean and well located?" has no valid answer for a spotless venue in a bad place. Split anything containing "and" or "or" into separate questions and the ambiguity disappears.

Keep the wording neutral

Bias creeps in through adjectives and framing. "How helpful was our excellent support team?" carries its own answer. So does offering unbalanced options — if your scale runs from "good" to "outstanding", you have not measured satisfaction, you have measured politeness.

• Balance your scales: equal positive and negative options, with a genuine neutral midpoint.
• Avoid absolutes like "always" and "never" — they push people toward the middle.
• Define vague terms. "Regularly" means different things to different readers; "more than once a week" does not.
• Put demographic questions last, unless you need them to route people to the right section.

Choose a scale and keep it

Five-point and seven-point scales both work. What does not work is mixing them, or changing the direction partway through so that agreement is on the left in one question and the right in the next. Consistency is what makes answers comparable, and comparability is the entire point of a questionnaire over a conversation.

Decide about required questions deliberately

Making everything mandatory looks rigorous and produces worse data. A respondent who cannot honestly answer a required question either abandons the questionnaire or picks something arbitrary — and an arbitrary answer is more damaging than a blank one, because you cannot tell it apart from a real one. Require only what you genuinely cannot analyse without.

Pilot it on five people

Give the draft to five people who resemble your intended respondents and watch them answer it. Every questionnaire has at least one question that is obvious to its author and ambiguous to everyone else, and this is the cheapest way to find it. Ask them to narrate what each question means to them as they go — the mismatches surface immediately.

Structure long questionnaires into sections

Grouping related questions into short, titled pages makes a long questionnaire feel manageable and gives a natural place for progress. It also lets you use logic: someone who says they have never used a feature should not then face six questions about it.