How to Write Survey Questions

A study's quality is often tied to sample size or analysis methods. But the quietest and deepest threat to results is poorly written questions. A badly constructed question pushes respondents toward a certain answer, confuses them, or measures something you didn't intend. In this article, we explain what a good survey question looks like, when each question type works best, and the most common traps to avoid.

What makes a good survey question?

A good question is one the respondent understands on first reading and can answer without struggle. Sounds simple, but most questions fail this test in practice.

Measures one thing

Doesn't squeeze two topics into one question. Each question measures only one dimension.

Means the same to everyone

Free of jargon, vague expressions, and ambiguous words.

Doesn't lead the answer

Doesn't push respondents toward a particular response. Uses neutral language.

Asks something knowable

Asks things respondents can realistically remember or know.

These four qualities are interconnected. A question not focused on a single topic tends to be leading too. A question that doesn't mean the same to everyone produces inconsistent responses.

Question types and when to use them

Choosing the right question type is as important as writing the question itself. Each type has situations where it's strong and where it falls short.

Closed-ended questions

Respondents choose from predetermined options. Single-select for 'which one most,' multi-select when multiple answers are valid. Also great for ad tests, packaging tests, and message tests where you need clear comparable results.

Visual/video-enabled questions

Show an image, video, or audio file then ask evaluation questions. The most direct way to test ad creatives, social media visuals, or packaging designs before going live.

Open-ended questions

Respondents answer in their own words. Valuable for unexpected insights but laborious to analyze. Best used at a few critical points rather than throughout.

Scale questions (Likert)

Measures degree of an attitude (strongly agree - strongly disagree). Frequently used in satisfaction, intent, and perception measurements. 5-point or 7-point scales are most common.

Ranking questions

Asks respondents to rank options by importance or preference. Useful for understanding relative priorities. Best to keep at 5-6 items maximum.

Matrix questions

Evaluates related items on the same scale. Produces comparable data but risks 'straight-lining' if too many rows are included.

NPS questions

'How likely are you to recommend on a 0-10 scale?' Simple, comparable, and trackable over time. Usually followed by an open-ended 'why' question.

Core rules for writing good questions

A few simple rules save the vast majority of questions. None are complex but all are easily forgotten.

Ask one thing per question

'Are you satisfied with our price and quality?' is a classic double-barreled question that produces uninterpretable data. Split every such question.

Avoid leading language

'How much did you enjoy our quality products?' assumes quality from the start. Neutral formulation is always more reliable.

Ask what respondents can know

'How much did you spend in this category last year?' asks something most can't accurately recall. It generates fabricated answers.

Use clear, simple language

Jargon and long sentences tire respondents. Write the question as if asking someone who knows nothing about the topic.

Keep options balanced and comprehensive

Overlapping ranges confuse respondents. Add 'Other' or 'None' options when needed to fill gaps.

Common mistakes

The most frequent mistakes in survey design and how to avoid them:

Social desirability bias

'Do you prefer eco-friendly products?' Almost everyone says yes. To measure real behavior, frame questions around concrete actions rather than intentions.

Too many options

With 15 options, respondents don't read them all—they pick the first few. Randomizing option order reduces this primacy effect.

Questions with built-in assumptions

'How often do you recommend our brand?' assumes they already do. Ask whether they recommend before asking frequency.

Making the survey too long

In long surveys, respondents fatigue and start answering randomly. Before adding each question, ask 'which decision will this answer inform?'

Negative or double-negative phrasing

'Wouldn't you not recommend this service?' confuses respondents. Keep questions positive and straightforward.

Question order matters too

Writing each question correctly isn't enough—their sequence also affects results. A preceding question can influence the answer to the next. This is called 'order effect.'

For example, asking about positive experiences with a brand right before asking overall satisfaction inflates the score, because the respondent's mind just focused on positive memories.

Placing general and broad questions first, specific and detailed ones later typically creates a safer flow. Sensitive or personal questions go last, as respondents are more willing to answer them after warming up to the survey.

How Sorbunu makes question design easier

Writing good questions takes experience, but you don't have to start from scratch every time.

Sorbunu offers 13 different question types ready to use, each designed for a specific measurement need. From single and multi-select closed-ended questions to scale and matrix questions, NPS to visual/video-enabled questions.

The platform is structured to prevent common design mistakes from the start. Randomizing option order, separating screening questions, and tracking survey length are built into the flow.

After designing your questions, matching them with the right sample size is possible within the same flow. A good question set combined with adequate, targeted sample becomes truly decision-driving research.

Frequently Asked Questions

No fixed number, but the general rule is to keep it as short as possible. In consumer research, 10-15 questions cover most needs. In longer surveys, respondent fatigue kicks in and data quality drops for later questions.

They serve different purposes. Closed-ended for clear comparisons and fast analysis. Open-ended for unexpected insights and respondents' own language. The most efficient approach is mostly closed-ended with open-ended at a few critical points.

5-point and 7-point are most common. 5-point is easier for respondents and sufficient in most cases. 7-point allows finer distinctions. Even-numbered scales (4 or 6) remove the neutral midpoint, forcing respondents to lean one way.

Simple test: if someone reading the question can sense which answer is 'expected,' it's probably leading. Positive adjectives, assumptions, and emotional language create bias. Rewriting in neutral, descriptive language usually fixes it.

Screening questions determine who participates in the research and typically appear at the start. Regular questions cover what you actually want to measure. Proper screening prevents collecting data from people outside your target audience.

Yes, often more than expected. A preceding question can frame the respondent's mind and influence subsequent answers. A general-to-specific flow, keeping sensitive questions last, and separating potentially influencing questions reduces this risk.

Build Your Survey with the Right Questions

Choose from 13 question types on Sorbunu, reach your target audience, and go to field in minutes.

Try for Free