TL;DR:
Decide what you will do with each answer. Cut every question you could not act on.
One idea per question, plain words, and a mix of quick choices and one or two open boxes.
Tell people how long it takes, show progress, and test it on someone before sending.
A survey with thirty questions gets abandoned at question twelve, and the answers you most wanted are at the end. People are being kind enough to help you. The way to respect that is to ask less, ask clearly and say how long it will take.
Start from the decision
Before writing a question, write down the decision the survey will inform. For example: "Should we keep the Saturday opening hours?" Then every question must connect to a decision. If you cannot say what you would change depending on the answer, delete the question.
Keep a small number of questions. It is easier to run a second short survey later than to fix a long one that nobody finished.
Before you start
Write the decision and the audience in a sentence each.
Choose your tool and check what question types it has, rather than assuming.
Decide where answers will be collected so you can read them. A form that sends answers to a spreadsheet is easy to total.
Plan to test it on one or two real people.
Step by step
List possible questions: Write everything you are curious about, then cut to the ones tied to the decision.
Make each question about one thing: "How was the speed and the price?" is two questions. Split it.
Choose the answer type: Use a scale or a choice for things you want to count, and an open box for "why".
Order from easy to hard: Start with a simple, interesting question. Put personal questions last.
Add a short intro: What it is for, how long it takes, and whether answers are anonymous.
Test it: Ask one person to fill it in while you watch. Fix where they hesitate.
Worked example: a ten-question customer survey
A small shop wants to know whether to keep Saturday opening hours. The survey has ten questions.
# | Question | Type |
|---|---|---|
1 | How often did you visit in the last month? | Choice |
2 | Which day do you usually visit? | Choice |
3 | How easy was it to find what you needed? | Scale of 1 to 5 |
4 | How would you rate the helpfulness of staff? | Scale of 1 to 5 |
5 | How would you rate our opening hours? | Scale of 1 to 5 |
6 | Would later weekday hours help you more than Saturday hours? | Yes, No, Not sure |
7 | What time do you usually arrive? | Choice |
8 | How did you first hear about us? | Choice |
9 | What is one thing we should change? | Open box |
10 | Is there anything else you would like to tell us? | Open box (optional) |
Check that every question has a purpose. Questions 2, 5, 6 and 7 bear on the opening hours decision directly. Questions 3 and 4 give context. Question 1 filters regulars from one-off visitors. Question 8 is about marketing and could be cut. If you do not need it, cut it and keep nine.
Now estimate the time with an assumption. Say a choice or scale takes about 15 seconds and an open box about 60 seconds. Then 8 closed questions × 15 s = 120 s, and 2 open boxes × 60 s = 120 s, a total of 240 s, which is 4 minutes. So you can honestly write "about 4 minutes". Your own timing from the test is a better number than this assumption.
Notice also that question 6 mixes two ideas in a mild way (later weekday hours versus Saturday hours). Rewrite it as two questions if you need precision: "Would you visit on a weekday evening?" and "Would you visit on a Saturday?" The answers can then be compared.
Wording that gets honest answers
Avoid leading phrasing. "How great was our service?" pushes the answer. Ask "How would you rate our service?"
Give every choice question an "other" or "not sure" if people may not fit your list.
Label the ends of a scale: 1 is very poor, 5 is very good.
Use the words your audience uses, not internal terms.
Progress cues and length
People continue when they can see the end. If your tool offers a progress bar or a page count, turn it on. Otherwise say "10 questions" in the intro and number them. Keep the whole thing on as few screens as possible. Make only the questions you truly need required, because a forced answer to an awkward question often produces a random click.
Mistakes to avoid
Asking everything you ever wanted to know.
Making all questions required.
Sending without a test.
Promising anonymity and then asking for a name.
Collecting answers and never telling people what changed.
Doing it faster
While you build the form, DuckyHelper can look at your screen when you press the key and read your questions back out loud, which is a useful way to hear the double-barreled ones. The decisions stay with you.
Frequently asked questions
How long should a survey be?
As short as the decision allows. Ten questions that take a few minutes is a sensible target for a customer survey.
Should I use open questions?
One or two, placed near the end, because they take longer to answer and to read.
Get started
