Surveys are a common source of data about attitudes and behaviour. Their value depends almost entirely on design.
Define the Purpose
Know what decisions the survey will inform and what you need to learn. Every question should earn its place; long surveys lower response rates and quality.
Writing Good Questions
- Use plain, specific language.
- Ask one thing per question — avoid "Was the service fast and friendly?"
- Avoid leading wording: "How much did you enjoy…" presumes enjoyment.
- Provide balanced answer options, including "don't know" or "not applicable" where appropriate.
- Ask about specific, recent behaviour rather than general habits.
Scales
Rating scales (such as five or seven points from "strongly disagree" to "strongly agree") should be consistent across questions, with labelled points.
Order and Flow
Start with easy, relevant questions; put sensitive and demographic questions later; watch for earlier questions influencing later answers.
Test First
Pilot with a small group, ask them what they understood each question to mean, and revise.
Sampling and Response
Consider who receives the survey and who responds. Low or skewed response rates limit what you can conclude; weighting helps only partly.
Analysis
Report the number of respondents and response rate, show the distribution of answers rather than only averages, and be cautious with small subgroups. Free-text answers are valuable; language models can help summarise them, with human checking.
Respect Respondents
Be clear about how data will be used, keep responses confidential as promised, and don't collect more personal data than needed.