What you need to do before you write a single question
Before you open a survey tool, decide what you actually want to know and who you want to ask. A survey that tries to answer five different questions at once confuses both you and the person filling it out. Write down your main goal in one sentence — for example, "I want to know whether my customers prefer email updates or text messages" or "I need to understand which features my users find most confusing."
Next, think about who will take your survey. Are you asking your existing customers, random people on the internet, employees, or members of a specific group? The people you ask should be the people who can actually answer your question. If you want to know how parents feel about a product, asking teenagers will not give you useful information.
Finally, decide how many responses you need. A survey with five responses tells you almost nothing. A survey with fifty responses from the right group of people tells you something real. You do not need thousands of responses for most purposes, but you do need enough that one person's unusual answer does not skew your results.
Key Takeaways
- Write down your main question before you start building, so every question in your survey points toward answering it.
- Choose a survey tool that matches what you need — Google Forms for simple surveys, Typeform or SurveySparrow for something more polished, Qualtrics if you need advanced logic.
- Keep your survey short: five to ten questions is usually enough, and longer surveys get fewer complete responses.
- Test your survey yourself and with one or two other people before you send it out, so you catch confusing wording and broken links.
- When you share your survey, be clear about how long it takes and what you will do with the answers.
Choosing the right tool for what you are building
Google Forms is free and works well if you need something straightforward. You can create multiple-choice questions, open-ended text boxes, rating scales, and checkboxes. Responses go into a Google Sheet automatically, which makes it easy to see patterns. The downside is that Google Forms looks basic and does not let you show different questions to different people based on their earlier answers.
Typeform and SurveySparrow are paid tools that let you build surveys that look more polished and professional. Both let you show questions one at a time instead of all at once, which feels less overwhelming to the person taking the survey. They also let you use conditional logic — meaning you can show question B only if someone answered "yes" to question A. Typeform has a free tier with limited responses; SurveySparrow charges per month.
Qualtrics is expensive and built for research teams and large companies, but it handles complex surveys with branching logic, randomized question order, and detailed analysis. Unless you are running a formal research study, you probably do not need it.
For most people, Google Forms is enough to start. If you find yourself wishing your survey looked nicer or that you could skip questions based on answers, move to Typeform or SurveySparrow.
Writing questions that actually get you answers
The most common mistake is asking two things in one question. "Do you like the product and would you recommend it?" is two questions. Someone might love the product but not want to recommend it to their friends. Split it into two separate questions.
Avoid leading questions that push people toward a particular answer. "Don't you think our service is excellent?" is leading. "How would you rate our service?" lets people give you their honest opinion. The same applies to answer choices — if you offer "Excellent," "Good," and "Okay," you are nudging people toward the positive side. If you want honest feedback, offer a balanced scale like "Very satisfied," "Somewhat satisfied," "Neutral," "Somewhat dissatisfied," "Very dissatisfied."
Keep the language simple and specific. "What is your experience with our platform?" is vague. "How often do you use our platform each week?" is specific and answerable. People will give you better data when they know exactly what you are asking.
For open-ended questions where people type their own answer, keep them optional. Some people will write paragraphs; others will skip them. That is fine. Required open-ended questions often get one-word answers or people abandon the survey entirely.
Building your survey step by step
Start with a brief introduction that explains what the survey is for and how long it takes. "This survey takes about three minutes and helps us understand which features you use most" is better than no introduction at all. If the survey is anonymous, say that. If you are collecting names or email addresses, say what you will do with them.
Put your most important questions in the middle, not at the beginning or end. People are freshest in the middle of a survey. The beginning is where they are still deciding whether to keep going, and the end is where they are tired and rushing.
Group related questions together. If you are asking about customer service, ask all the customer service questions in a row. Jumping between topics makes the survey feel disorganized and confusing.
End with demographic questions like age, location, or job title if you need them. People are more willing to answer personal questions at the end, after they have already invested time in the survey. If you ask for personal information at the start, some people will quit immediately.
Testing your survey before you send it out
Take your own survey from start to finish. Does every question make sense? Are the answer choices clear? Does the survey flow logically? Write down anything that confused you or took longer than expected.
Ask one or two other people to take it and watch them do it if you can. Do not explain the questions — just let them read and answer. If they ask for clarification, that means your question needs to be clearer. If they misunderstand what you are asking, rewrite it.
Check that your links work if you included any. Check that the survey looks right on a phone as well as a computer — many people will take it on their phone. If you are using conditional logic, test both paths: answer "yes" to the branching question and make sure the right follow-up appears, then go back and answer "no" and check that a different question appears.
Sharing your survey and collecting responses
The way you share your survey affects who takes it. Sending it to your email list reaches people who already know you. Posting it on social media reaches a broader group but also gets more random responses. Paying to promote it on Facebook or Google reaches a specific audience you choose, but costs money.
Be transparent about what you are doing with the data. "We are using this to decide which features to build next" is honest. "We will never sell your information" is reassuring if that is true. People are more likely to take a survey when they understand why you are asking.
Send a reminder after a few days if you are sending to a specific group. Many people will not see the first message or will mean to take it later and forget. A second message usually increases your response rate by 20 to 40 percent.
Set a deadline so you know when to stop collecting responses and start analyzing. Two weeks is typical for most surveys. After that, responses usually slow down and the people still answering may be different from the people who answered early.
Making sense of what you collected
Most survey tools will show you basic numbers automatically — how many people chose each answer, what percentage that represents. Look for patterns. If 70 percent of people chose the same answer, that tells you something clear. If responses are split evenly, people disagree and you may need to dig deeper.
For open-ended questions, read through the answers and look for themes. You do not need to read every single one if you have hundreds of responses, but read enough to see what people are saying. Write down the most common complaints or suggestions.
Compare answers between groups if you collected demographic information. Maybe younger users feel differently than older users, or people in different regions have different needs. That comparison is often more useful than the overall numbers.
Remember that a survey tells you what people say, not always what they do. Someone might say they want a feature but never use it if you build it. A survey is a starting point for understanding, not a final answer.
Frequently Asked Questions
How many questions should my survey have?
Five to ten questions is usually enough. Longer surveys get fewer complete responses because people run out of patience. If you have more than ten questions, you probably need to narrow down what you actually want to know. Every question should point toward answering your main goal.
Should I make all questions required or let people skip them?
Making questions required means you get complete data, but it also frustrates people who do not want to answer something. A middle ground is to make most questions required but leave open-ended questions optional. People are more willing to skip a question where they have to type their own answer.
What is the difference between a rating scale and a ranking question?
A rating scale asks people to rate each item separately — "Rate each feature from 1 to 5." A ranking question asks people to put items in order — "Rank these features from most to least important." Ranking is harder for people to do and works better when you have fewer than five items. Rating scales are easier and more common.
How do I know if my survey results are reliable?
The more responses you get from the right group of people, the more reliable your results are. Fifty responses from your actual customers tells you something real. Five responses tells you almost nothing. If your results are split 51 percent to 49 percent, that difference might just be random chance. If they are 80 percent to 20 percent, that is a real pattern.
Can I change my survey after people have started taking it?
You can, but it makes your data messy. Some people will see the old version and some will see the new version, and you will not know which is which. If you need to make changes, close the survey, make them, and reopen it. Tell people you made changes so they know to take it again if they already did.