GuideData & AI
Designing a survey that produces analysable data
By Inara Crest Engineering 1 min read
Clear objectives, well-formed questions, a sound sample and a pilot run turn survey responses into evidence you can analyse.
A survey can only be analysed as well as it was designed. Decisions made before the first response is collected determine what you can conclude afterwards.
1. Write down the questions you need answered
List the research questions and the analysis you plan for each — a comparison between regions, a trend over time, a relationship between two factors. Every survey question should serve one of them.
2. Write clear, neutral questions
- Ask about one thing per question.
- Avoid leading wording ("How satisfied are you with our excellent service?").
- Offer balanced answer scales with a clear midpoint where appropriate.
- Include "don't know" or "not applicable" options when they are genuine answers.
3. Plan the sample
Define who the survey represents and how respondents will be selected. A large sample from the wrong population is less useful than a smaller, well-chosen one.
4. Choose the collection method
Digital forms on phones or tablets reduce transcription errors, validate answers as they are entered and can record location and time where that is relevant and consented.
5. Pilot before launch
Test the survey with a small group, time it, and check that the data exports in the structure you need for analysis.
Planning a study? See our research and data collection services.
About the author
Inara Crest Engineering
Engineering team
Written by the Inara Crest Technologies engineering team, who design, install and support software, network, security, power and data centre systems for organisations in Kenya and East Africa.