How to Make a Survey: 9 Steps From Goal to Action

Knowing how to make a survey means turning a vague question into a decision. You pick a goal, ask the right people a few clear questions, and plan what you will do with the answers. A survey is a structured set of questions sent to a defined group to measure attitudes, behaviors, or experiences.

Most surveys fail before launch, and the cause is rarely the software. The goal is fuzzy, the audience is wrong, or the questions pull in two directions.

In this guide, we’ll walk through nine steps, from goal to action, with sample size math, question types, real examples, and the metrics that show whether your survey worked.

What is a survey, and how does it differ from similar methods?

A survey is a structured set of questions sent to a defined group to measure attitudes, behaviors, or experiences. An online survey is the same method delivered over the internet, through email, a web link, a QR code, or social media.

Surveys are standard practice. A study found that 99% of responding UX researchers run surveys at least sometimes. The terms below get mixed up often, so here is how they differ.

Choose a survey when your question is about how many or how much, and about what people say. Other methods, such as usability testing, for questions about why people behave as they do.

What are the pros and cons of online surveys?

Online surveys reach many people quickly and cheaply, but they trade depth and control for speed. Knowing both sides helps you decide whether a survey is the right tool.

Pros

• They reach large groups fast, across locations and time zones.
• They cost less than phone or in-person research.
• Results arrive in real time, so you can monitor progress.
• Logic, images, and video make the experience flexible.
• Respondents answer on their own time, with no interviewer to influence them.

Cons

• People who choose to respond may not represent your whole audience.
• Some groups, such as older adults, can be harder to reach online.
• Emails can land in spam or go unread.
• No one is present to clarify a confusing question.
• Rushed or careless answers can lower data quality.

How do you make a survey step by step?

You make a survey in nine steps: set a goal, define the audience, pick a platform, write the questions, build the structure, pilot it, distribute it, analyze it, and act on the results. The order matters, because each step narrows the next.

1. Set a goal tied to a decision

Start with the decision the survey will inform, not a topic. “Customer feedback” is a topic. “Decide whether to redesign onboarding” is a goal. A specific goal tells you which questions to ask and who should answer them.

Then plan the practical side. Set a budget, a deadline, and a benchmark for success, such as a target number of completed responses.

• Weak goal: Learn what customers think.
• Strong goal: Find out why trial users do not upgrade to a paid plan.
• Strong goal: Test three package designs before a retail launch.

2. Define your audience and sample size

Your sample is the group of people who actually answer. Choose people who match your goal, such as current customers, employees, or non-customers. If you cannot control the list, open with a screening question that filters out people who cannot answer from experience.

Then size the sample. This formula works for a known population: n = (Z² × p × q × N) / (N × e² + Z² × p × q). Here, n is the sample size and N is the population. Z is the confidence score, 1.96 for 95%. The margin of error is e. The values p and q are the expected split of answers, and 50% each is the safe default when you do not know.

These figures use the standard formula. For a population of 1,000 at 95% confidence and a 5% margin, you need 278 completes. Plan invitations around your expected response rate. At a 30% rate, 385 completes means inviting about 1,284 people.

3. Choose a platform, then a template

Pick the tool before you design, because tools differ in question types, logic, and reporting. The right choice depends on the size and repeat value of your project. Here is how the main options compare.

• Form tools such as Google Forms or Microsoft Forms: Best for short, one-off surveys with a small audience. Logic, question types, and reporting are limited. For a one-page form, see our guide on how to create a survey form.
• Spreadsheet or Word forms: Best for offline, paper-style collection. Expect manual data entry and no live results.
• Social media polls: Best for one-question pulse checks. You get no segmentation and a self-selected audience.
• Survey platform: Best for repeat, larger, or research-grade surveys. It has more features to learn, plus plan limits.

Features vary by product, so check current plans. When you compare platforms, look for varied question types with validation, skip logic, mobile-ready design, real-time reporting, data export, integrations with your CRM, language support, and clear security and privacy terms. A survey software platform should cover most of these. Start from a template if one fits your goal, then edit it.

4. Write questions that do one job each

Give every question one job. A double-barreled question asks about two things at once, such as “How satisfied are you with our pricing and support?” The answer cannot tell you which one the person meant. Split it in two. Keep wording neutral, plain, and short, and for deeper examples see our guide to good survey questions.

Match the question type to the answer you need:

• Closed-ended: Fixed options such as multiple choice, yes or no, and rating scales. Fast to answer and easy to analyze.
• Open-ended: Free text. It explains the why, but takes more effort.
• Likert scale: Agreement on a graded scale, such as strongly disagree to strongly agree.
• Net Promoter Score (NPS): A 0 to 10 rating of how likely someone is to recommend you.
• Matrix: A grid of items rated on one scale. Use it sparingly.
• Ranking: Ordering items by priority.

Mixing types usually gives richer data than using one alone.

5. Set the stage and structure the flow

A survey starts before the first question. Tell people what the survey covers, how long it takes, and whether answers are anonymous or confidential. Anonymous means you collect no identifying data. Confidential means you collect it but never share it. Mention any incentive here too.

• Order questions from easy to sensitive, and place demographics last.
• Use skip logic, a rule that sends people past questions that do not apply, so each person sees only relevant items.
• Randomize answer order where it is safe to do so, to reduce order bias.
• Design for phones. Grids encourage straightlining, where people pick the same answer down a column, and force scrolling on small screens.

6. Pilot test with 5 to 10 people

A pilot is a small trial run with people like your respondents. Ask five to ten testers to complete the survey on their own devices, ideally while you watch.

Check four things: how long it really takes, which questions confuse people, whether the logic routes correctly, and how the layout looks on a phone. Fix what you find. If you change a lot, test again with a few new people.

7. Distribute the survey and send reminders

Choose the channel your audience already uses. Then handle timing, reminders, and controls.

• Email: Best for known contacts. Use a real sender name and state the time needed.
• Web link or embed: Best for visitors on your site.
• QR code: Best for events and print.
• Text message: Best for mobile-first audiences, and only with their consent.
• Social media: Broad reach, with less control over who answers.
• Offline collection: Best where connectivity is poor.

Midweek mornings are a common starting point, but test your own audience. Our guide to the best time to send a survey explains how. Limit each person to one response, and send one or two reminders only to people who have not answered. Close with a short thank-you. Match any incentive to the effort, such as a gift card, a donation, or a prize draw.

8. Analyze results against your plan

Decide how you will analyze the data before responses arrive. Name the segments you will compare, the metrics you will track, and the charts you will build. That plan keeps you from fishing for patterns after the fact.

When responses come in, build reports and filter by segment, device, or location. Cross-tabs compare answers across groups. As a rule of thumb, wait for at least 30 responses in a segment before comparing it with another. For open-ended answers, word clouds show which terms appear most, and sentiment analysis tags text as positive, negative, or neutral.

9. Act on the findings and close the loop

Data without action is wasted effort. Turn each finding into a decision, an owner, and a date. A phone maker might use habit data to prioritize features. A bank might use knowledge gaps to refine its marketing.

Then tell respondents what you heard. People who see their input matter are far more likely to answer next time.

• Share the top findings in plain language.
• Name one or two changes you will make.
• Give a rough timeline for those changes.

What do real surveys look like in practice?

Real surveys share a pattern: one decision, one audience, and a few focused questions. The table shows four common use cases with an example goal and question for each.

Take the package test. Show each design to a sample of target buyers, ask for a first choice and a reason, and compare the results. You learn which design wins and why, before you spend on a full print run.

How do you measure whether your survey worked?

Judge a survey by five checks: response rate, completion rate, completion time, representativeness, and whether the data answered your original question. Together they show whether the right people finished and whether the results are usable.

1. Response rate.
   Divide completed surveys by eligible invitees and multiply by 100. If 400 of 1,000 invitees finish, the rate is 40 percent. Our guide to survey response rates shows the math.
   
2. Completion rate.
   Of the people who started, how many finished? A sharp drop at one question points to a problem there.
   
3. Completion time.
   Compare the median with the time you promised. As a rule of thumb, investigate if it runs 20 percent over.
   
4. Representativeness.
   Compare respondents with your target group on known traits such as age, region, or customer plan.
   
5. Goal fit.
   Check that the results answer the question you wrote in step 1.

What are the most common survey mistakes?

Most mistakes come from rushing the plan. For a longer checklist, see our guide to survey best practices.

• Leading or double-barreled questions: They push answers or blur them.
• Surveys that run too long: Every extra question lowers the odds of finishing.
• Skipping the pilot: Broken logic and unclear wording go live.
• Surveying the wrong audience: The data may be clean and still useless.
• Ignoring response bias: If only people with extreme opinions respond, the results skew.
• No analysis plan: You collect data you cannot use.
• Never closing the loop: People stop responding when they see no result.

How do you make a survey in QuestionPro?

QuestionPro offers three ways to start a survey: pick a template, import an existing questionnaire, or describe what you need. QuestionPro AI builds a draft from a plain-language description and suggests questions you can add, reorder, or edit. You can also import a Word file instead of retyping it.

Two features save time on steps 6 and 7. A preview mode shows the survey as respondents will see it, and you can generate random test responses to check how reports will look before launch. Quota controls cap responses per group to keep the sample balanced, and audio and video questions let people answer out loud or on camera. For field research, the mobile app collects responses offline. A free plan lets you try the workflow first.

Start with the decision, and the survey follows

Every choice in this guide traces back to one question: what will you do differently once you have the answers? A clear decision shrinks the audience, shortens the questionnaire, and makes the analysis obvious. Fancy software cannot replace that clarity.

Before you build your next survey, write down three things:

• The decision the results will inform.
• The group of people who can answer.
• The action you will take if the results go one way, and the action if they go the other.

If you can fill in all three, you are ready to start.

Frequently Asked Questions (FAQs)