What is Survey Design?
Survey design is the engineering of a questionnaire. Decisions span question type (open vs closed, multiple choice, Likert, semantic differential), scale (5-point or 7-point, balanced or unbalanced, anchored or not), order (general before specific, sensitive items late), wording (no double-barreled, leading, or jargon-laden questions), and length (under 12 minutes online to avoid drop-off). A good design yields high response rates, low non-response bias, and answers that map cleanly onto the original research question.
How Survey Design actually works
The framework breaks down into the following moving parts. Knowing what each piece is — and what it is not — is what separates a B-grade answer from an A-grade answer in a written assignment.
- Question type — closed for measurement, open for discovery, mix sparingly
- Scales — Likert for attitudes, semantic differential for image, ordinal for ranking
- Order — funnel from general to specific; demographics last
- Wording — one idea per question, no leading language, no jargon
- Pre-test — pilot with 20–30 respondents and revise before fielding
A worked example: A Net Promoter Score program
The Net Promoter Score uses one of the most studied survey questions in marketing: "How likely are you to recommend us to a friend or colleague?" on an 11-point scale. The wording is single-barreled (one idea), the scale is unambiguous (0 = not at all, 10 = extremely), and the placement is typically right after a transaction so the experience is fresh. Promoters (9–10) minus Detractors (0–6) equals NPS — a metric that survives because the question design holds up across cultures and categories.
Don't lose marks for these
- Double-barreled questions ("How satisfied are you with our quality and price?") — split them
- Leading questions ("Don't you agree our service is excellent?") — neutral phrasing only
- Asking sensitive questions early — drop-off goes up
- Using jargon the respondent will not understand
How to use this on the exam
Score-maximizing moves
- Identify at least three threats to validity in any survey scenario
- Recommend a pilot of 20–30 respondents
- Specify scale type (Likert, semantic differential) and justify the choice
When to use Survey Design (and when not to)
Use Survey Design when your assignment asks you to analyze, structure, or recommend — and when you have at least two data points to populate every cell of the framework. Skip it when the question is asking for a numerical answer or a single recommendation, since Survey Design is a structuring tool, not a calculator.