Start with the decision, not the questions
Before the first question, write down the one or two decisions the answers should inform. Every question that doesn’t help with them can go. Ten questions or fewer is a good target, and say how long the survey takes in the first line.
Build it
- 1 Start from a survey template, like Net Promoter Score or Team retrospective, or click Create a poll and add questions with Add content.
- 2 Use Rating or Opinion scale for how people feel, Matrix to rate several statements at once and Long text for the why.
- 3 Put closed questions first and open ones last. One easy question at the start gets people going.
- 4 Add a Page break every few questions and turn on the Progress bar.
- 5 Preview on a phone, then Publish.
Questions that give clean answers
- Keep them short, neutral and about one thing. “How satisfied are you with our fast and friendly support?” already suggests the answer.
- Ask about real experiences: “How often did you use…” beats “Would you use…”.
- Keep the same scale throughout, so 5 always means the same.
- Offer “Not applicable” or Allow "Other" answer where some people won’t fit your options.
- Make only the essential questions Required.
With Show only if under Logic, a follow-up only shows to people it applies to, like “What would have made it better?” after a low rating.
Get enough responses
Send the survey where people already are: the team chat, the order confirmation, the last slide of a meeting. Embed it on your site, or send it to an invite list and remind them two days before you close. Say why you ask and what will change because of the answers.
For anonymous surveys, leave out names, keep Show respondent names on Private and use unique codes instead of email addresses to keep it to one answer per person.
From answers to action
Results update live, and you can export every response as a CSV. Connect Slack, Discord, Google Sheets, Notion or Airtable to see responses where your team works. Responses are unlimited on every plan, Free included, so you can ask everyone rather than a sample. Then close the loop: tell people what changed.