Great research starts with great survey design. Even the most sophisticated analysis can't salvage insights from a poorly designed questionnaire. Whether you're a seasoned researcher or launching your first survey, these best practices will help you collect data that's accurate, actionable, and valuable.
The Foundation: Clear Objectives
Before writing a single question, crystallize your research objectives:
Ask yourself:
- What decisions will this data inform?
- What hypotheses are we testing?
- Who needs these insights and why?
- What actions will we take based on different findings?
Without clear objectives, surveys become unfocused wish lists of "nice to know" questions that exhaust respondents and dilute insights.
Question Writing Fundamentals
Keep It Simple
The best questions are easily understood by all respondents:
❌ Poor: "To what extent do you agree that the implementation of omnichannel retail strategies has positively impacted your propensity to engage with brands across multiple touchpoints?"
✅ Better: "How has shopping both online and in stores affected how you feel about this brand?"
One Concept Per Question
Double-barreled questions create ambiguous responses:
❌ Poor: "How satisfied are you with our product quality and customer service?"
✅ Better: Ask as two separate questions—one about product quality, one about customer service.
Avoid Leading Questions
Neutral wording prevents bias:
❌ Leading: "Don't you agree that our new packaging is more appealing?"
✅ Neutral: "How would you rate the appeal of our new packaging?"
Be Specific About Timeframes
Vague time references create inconsistent data:
❌ Vague: "How often do you exercise?"
✅ Specific: "In a typical week, how many days do you exercise for 30 minutes or more?"
Response Scale Best Practices
Choose Appropriate Scales
Match scale type to question type:
Agreement scales: Strongly disagree → Strongly agree
Satisfaction scales: Very dissatisfied → Very satisfied
Frequency scales: Never → Always
Likelihood scales: Not at all likely → Extremely likely
Decide on Scale Length
- 5-point scales: Good for most purposes, easy for respondents
- 7-point scales: More granularity, still manageable
- 10-point scales: Maximum differentiation, can feel arbitrary to respondents
- 11-point scales (0-10): Useful for NPS and familiar to respondents
Include a Midpoint (Usually)
Odd-numbered scales allow neutral responses. This is appropriate when:
- Genuine neutrality is possible
- You don't want to force a direction
Remove the midpoint when:
- You need respondents to lean one way
- Neutrality would mask important differences
Label All Scale Points
Labeled scales improve reliability:
❌ Less reliable: 1 ☐ 2 ☐ 3 ☐ 4 ☐ 5 ☐
✅ More reliable: Very dissatisfied ☐ | Dissatisfied ☐ | Neutral ☐ | Satisfied ☐ | Very satisfied ☐
Question Types and When to Use Them
Closed-Ended Questions
Single select: Best for mutually exclusive options
- "What is your primary reason for using this product?"
Multiple select: When respondents may choose several options
- "Which of the following features do you use? Select all that apply."
Rating scales: For measuring intensity or degree
- "On a scale of 1-10, how likely are you to recommend us?"
Open-Ended Questions
Use sparingly but strategically:
Best for:
- Understanding "why" behind ratings
- Capturing feedback in respondents' own words
- Discovering issues you didn't anticipate
Placement tips:
- Place after related closed questions
- Limit to 2-3 per survey
- Make optional when possible
Questionnaire Structure
The Opening
First impressions matter:
- Start with engaging, easy questions
- Build rapport before sensitive topics
- Screen for qualification early
- Set expectations for length
The Middle
Organize logically:
- Group related topics together
- Progress from general to specific
- Place demanding questions when attention is highest (early-middle)
- Use transitions between sections
The Ending
Close thoughtfully:
- Move demographic questions to the end
- Include an open-ended feedback question
- Thank respondents
- Avoid important questions at the very end (fatigue zone)
Skip Logic and Branching
Smart surveys adapt to respondents:
Benefits:
- Shorter surveys for each respondent
- More relevant questions
- Better data quality
- Improved completion rates
Best practices:
- Test all logic paths thoroughly
- Don't create overly complex branching
- Ensure logical flow regardless of path
- Consider mobile display of branched questions
Survey Length and Timing
The Length Tradeoff
Every question has a cost:
- Longer surveys = lower response rates
- Longer surveys = higher dropout rates
- Longer surveys = degraded data quality (speeding, straightlining)
Guidelines
Reducing Length
- Cut "nice to know" questions ruthlessly
- Eliminate redundant questions
- Use matrix questions carefully (they feel longer)
- Consider splitting into multiple surveys
Mobile Optimization
Over 60% of survey responses now come from mobile devices:
Design for Small Screens
- Use single-column layouts
- Keep questions and responses visible without scrolling
- Limit matrix question size
- Use larger touch targets
Test on Mobile
- Complete your survey on a phone
- Check all question types display correctly
- Verify skip logic works
- Time the mobile experience
Pre-Testing Your Survey
Never launch without testing:
Internal Review
- Have colleagues complete the survey
- Check for typos, logic errors, broken paths
- Verify timing estimates
Cognitive Interviews
- Watch 5-10 people complete the survey
- Ask them to think aloud
- Identify confusing questions
- Note where they hesitate
Soft Launch
- Field to a small sample first (50-100)
- Check data quality and distributions
- Identify problem questions
- Make final adjustments
Data Quality Controls
Build quality checks into your design:
Attention Checks
"Please select 'Strongly agree' for this question" catches inattentive respondents.
Consistency Checks
Ask related questions differently and compare—major inconsistencies indicate poor quality.
Speeding Detection
Flag responses completed impossibly fast.
Open-End Quality
Review open-ended responses for gibberish or off-topic answers.
Common Mistakes to Avoid
Asking questions you won't use: Every question should tie to an objective.
Assuming knowledge: Don't ask about concepts respondents may not know.
Forgetting "Don't know" and "N/A": These options prevent forced guessing.
Overlapping response options: Ensure mutual exclusivity in single-select questions.
Recency bias in lists: Randomize option order when appropriate.
Conclusion
Survey design is both science and art. The principles are clear—simplicity, neutrality, logical flow—but applying them requires judgment and practice. Invest time in design, test thoroughly, and continuously improve based on data quality feedback.
A well-designed survey respects respondents' time, captures accurate data, and delivers insights that drive better decisions. That's worth the extra effort.