Design & Analyze a Survey

Use this when you need numbers from many people, not quotes from a few. Write the decision first, then the questions, then field, clean, and report.

Use this

  • A decision needs counts, not only stories
  • You already know the topics from interviews or support data
  • You can reach enough of the right people to trust the split

Skip this

  • You do not know what to ask yet. Interview first
  • You need to watch someone use a flow. That is a usability test
  • You are about to ship a change and have traffic. That is an A/B test

Steps

Goals and questions sit together. Sampling, pretest, and send are one fielding job. Clean before you analyze.

01Goals and questions

02Field the survey

03Clean and analyze

Frequently asked questions

How long should the survey be?

Five to ten minutes on a phone. Cut questions that will not change a decision. If you need stories, interview instead of stuffing open boxes at the end.

Do I need a hundred responses?

For a split by segment, yes, around there. For a directional pulse, fewer is fine if you do not treat it like a census. If you cannot reach the right people, do not field it.

We already did interviews. Do we still survey?

Yes, when you need counts on questions you already know. No, if you are still guessing what to ask. That is still a study.