How to Design a Research Survey: Likert Scales, Questions and Reliability

Surveys are the most common data collection tool in quantitative research. But a poorly designed survey gives unreliable results no matter how many people answer it.

First, check for an existing scale

If a validated scale exists for the concept you want to measure, using it is usually the best choice. Using someone else’s scale usually requires written permission. If the scale is in another language, check whether a validated adaptation exists.

Likert scales

Participants indicate how much they agree with a statement.

• 5-point: Strongly disagree – Disagree – Neutral – Agree – Strongly agree
• 7-point: allows finer distinctions.

Reverse-coded items

Some items are worded in the opposite direction ("I often feel inadequate."). They help detect participants who tick the same option without reading. They must be reverse-coded before analysis (on a 5-point scale, 1→5, 2→4).

Rules for writing good questions

• Ask one thing. "The course content was interesting and useful" is two questions.
• Do not lead. Openings like "Considering the benefits of online learning…" bias answers.
• Use plain language. Avoid jargon participants may not know.
• Avoid negatives and double negatives.
• Make options exhaustive and mutually exclusive. Age ranges must not overlap; add "Other" if needed.

Demographic questions

Ask about gender, age, year or department only if you will actually use them in the analysis. Collecting unnecessary personal data is an ethical and legal problem.

Consent and ethics

Start the survey with an informed consent text explaining the purpose, duration, voluntary participation and how data will be used. Research with human participants usually requires ethics approval before data collection.

Pilot testing

Try the survey with a small group before the main study to catch unclear items, timing and technical issues.

Reliability: Cronbach’s alpha

Alpha shows whether the items of a scale measure the same construct consistently. Values of .70 and above are widely considered acceptable, although interpretation depends on the purpose and number of items. Calculate and report reliability in your own sample too.

Surveys with Academiana

Academiana’s survey builder supports single and multiple choice, 5- and 7-point Likert scales (with reverse-item marking), numeric and open-ended questions. Add a consent text and share the survey via a link; no names, emails or IP addresses are collected from participants. Download responses as a spreadsheet or send them to your project in one click — reverse items are recoded automatically, so you can run analyses such as Cronbach’s alpha right away.