How to Create a Survey: Survey Design Steps, Question Types and Examples

Anyone can build a survey in a few minutes. Building one that people finish, that gives honest answers and that leads to a decision is harder. Most bad surveys fail for the same reasons: too many questions, leading wording, the wrong scale, bad timing, and nobody reading the open answers.

This guide covers survey design step by step, the main question types and when to use each, the biased questions to avoid (with better versions), how to use open-ended questions, example surveys you can adapt, and how to analyse the results.

What makes a good survey?

• One clear goal. Every question serves a decision you will make.
• Short. Ask only what you will use. Every extra question costs completions and attention.
• Neutral wording. Questions do not suggest the "right" answer.
• The right moment. Asked right after the experience, in the channel where it happened.
• A plan for the answers. Someone owns the results and acts on them.

How to create a survey, step by step

Step 1: Write down the goal

Finish this sentence before you write a single question: "We are running this survey so that we can decide ___." For example: "decide whether to change our delivery partner" or "find out why new customers do not renew". If you cannot finish the sentence, you are not ready to write questions.

Step 2: Choose who you ask

Define the audience: all customers, customers who contacted support this week, users of one feature, people who cancelled. Ask the people who actually had the experience. If you survey a sample, make sure it is not limited to your happiest or most active customers.

Step 3: Pick the moment and the channel

Timing changes answers. Ask about a delivery right after it arrives, about support right after a case closes, about onboarding at the end of the first week. Use the channel the customer is already in: on the website or in the app, by email, by SMS or WhatsApp, or by phone.

Step 4: Write the questions

Start from the decision and work backwards: what do you need to know to make it? For each piece of information, choose the right question type (see the table below) and write the question in plain, neutral language.

Step 5: Order and trim

Put the most important question first, so even partial answers are useful. Go from general to specific. Keep sensitive or demographic questions for the end, and only if you need them. Then remove every question that does not change a decision.

Step 6: Test it

Send the survey to three to five people outside the team. Watch where they hesitate, which questions they misread and how long it takes. Fix confusing wording and broken logic before launch.

Step 7: Send and remind

Send it, and if it is a one-off survey (not a moment-based one), send one polite reminder to people who have not answered. Tell people how long it takes, and keep that promise.

Step 8: Analyse and act

Look at scores by segment, read and group the comments, find the main reasons behind low scores, and decide what to change. Then tell respondents what you changed. That is what makes them answer next time.

Survey question types and when to use them

For metric-specific questions, see our guides to CSAT and NPS.

Tips for rating scales

• Label the ends, and ideally every point, so the meaning is clear.
• Keep the scale balanced: as many positive options as negative ones.
• Keep the same scale across surveys you want to compare.
• Decide on a neutral midpoint on purpose. Include one when "neither" is a real answer.

Biased survey questions to avoid

Small wording choices push people towards an answer. These are the most common types, with better versions.

Read every question as if you were a slightly annoyed customer. If it sounds like the company wants a particular answer, rewrite it.

Open-ended questions

Rating questions tell you what happened. Open-ended questions tell you why. A survey with no open question leaves you guessing about the reasons behind a score.

Good open questions:

• What is the main reason for your score?
• What could we have done better?
• What almost stopped you from buying?
• What would you change if you were in charge of [product]?
• Is there anything else you would like to tell us?

How to use them well:

• One or two per survey, usually right after the key rating question.
• Make them optional, so people who do not want to write still finish.
• Read them systematically. Group answers by theme and track how often each theme appears and how negative it is. With hundreds of answers a month, use text analytics; our sentiment analysis guide explains how.

How long should a survey be?

As short as the goal allows. A survey asked after a single interaction often needs only one rating question and one open question. A deeper research survey can be longer, but every question should still earn its place. Tell people how long it takes, show progress, and remove questions whose answers nobody will use.

How to get more people to respond

• Ask at the right moment, right after the experience, while it is fresh.
• Use the channel they are in: in the app, on the site, in the chat, by email or by phone.
• Keep it short and say how long it takes.
• Make it easy on mobile. Large buttons, one question per screen for longer surveys.
• Explain why you ask and what you will do with the answers.
• Do not over-survey. Limit how often the same person is asked.
• Close the loop. People answer again when they see that their feedback changed something.

Example surveys you can adapt

After a purchase

1. How satisfied are you with your purchase experience? (1–5)
2. What is the main reason for your rating? (open)
3. Was there anything that almost stopped you from completing your order? (open, optional)

After a support case

1. How satisfied were you with the help you received? (1–5)
2. Was your issue fully resolved? (yes / no)
3. What could we have done better? (open)

Relationship survey (every few months)

1. How likely are you to recommend us to a friend or colleague? (0–10)
2. What is the main reason for your score? (open)
3. What is one thing we should improve? (open)

Product or feature feedback

1. How useful is [feature] for your work? (1–5)
2. What do you use it for? (open)
3. What would make it more useful? (open)

After cancellation

1. What is the main reason you are leaving? (single choice, with "other")
2. What could we have done to keep you? (open)

Common survey design mistakes

• No goal. Collecting "feedback" without a decision in mind leads to reports nobody uses.
• Too many questions. Long surveys lose people halfway and tire the rest.
• Only rating questions. You learn that something is wrong, but not what.
• Leading or double questions. They produce answers you cannot trust.
• Asking too late. People do not remember the details of an experience from weeks ago.
• Surveying the same people too often. Fatigue lowers response rates and quality.
• Not reading the comments. The most useful part of the survey stays unused.
• Not acting or not telling customers. People stop answering surveys that change nothing.

How to analyse survey results

1. Clean the data: remove test answers, duplicates and obvious spam.
2. Calculate the key scores (CSAT, NPS, CES or your own) and compare them over time.
3. Break results down by segment: channel, product, region, customer type. Averages hide problems.
4. Group the open answers by theme and measure how often each theme appears and how negative it is.
5. Find the root causes of the biggest negative themes before deciding on fixes. See our guide to root cause analysis in customer feedback.
6. Decide, assign and follow up: each finding gets an owner and a date, and you check the scores again afterwards.

For a wider view of collecting feedback beyond surveys, see how to collect customer feedback.

Where Pivony fits

Sonic Survey is Pivony's survey tool, with a free tier of 25 responses a month:

• Build branded surveys with the usual question types, and put them where customers are: embedded on your site, as a pop-up, as a shareable link or QR code.
• Trigger them at the right moment, for example on exit intent, after a purchase, after a form or, on Pro and above, after a CRM or order event.
• Ask by voice on Pro and above: an AI agent can call customers or talk to visitors on your site, ask your questions and record the answers.
• Let AI read every answer: sentiment and themes on the free tier, and root cause analysis with the full VoC platform.

Next step

Try Sonic Survey for free, or request a free 24h root cause scan on a sample of your existing survey comments.