Likert scale survey questions: design, examples and coding

A Likert item asks respondents to place an attitude on an ordered set of labelled options; a Likert scale combines several items intended to measure the same underlying construct. Use one clear idea per item, match the labels to the construct, keep direction consistent and define coding and analysis before fieldwork.

What is a Likert scale?

A Likert item presents an ordered response continuum, commonly agreement, frequency, importance, satisfaction or likelihood. A multi-item Likert scale combines responses from several related items only when those items were designed and evaluated to measure one construct. A five-point agreement item alone is not evidence that a scale is reliable or valid.

Response-scale decision matrix

This matrix is a design aid, not a universal standard. Specify the construct first, then write the stem, anchors, point count, direction and non-substantive options as one testable unit.

Should you use five or seven points?

Five points are compact and easy to explain. Seven may be useful when the audience can make finer distinctions and the construct supports them. More categories do not automatically produce better data. Choose with cognitive testing, mobile display, comparability and the analysis plan in mind. Current SurveyMonkey guidance similarly treats five to seven points as a common range while emphasizing labelled options and audience context.

Should the scale include a midpoint?

Include a midpoint when a genuinely neutral position is meaningful. Remove it only when the research question justifies a directional choice and respondents can reasonably hold one. Never turn the midpoint into a catch-all for uncertainty or inapplicability; those states need their own options when relevant.

Twelve reusable item patterns

1. The instructions were easy to understand. — Strongly disagree to strongly agree.
2. I could complete the task without help. — Strongly disagree to strongly agree.
3. The result matched what I needed. — Strongly disagree to strongly agree.
4. How often did the error occur during the past 30 days? — Never to every day.
5. How often did you use the feature during the past 30 days? — Never to every day.
6. How satisfied are you with the time required? — Very dissatisfied to very satisfied.
7. How satisfied are you with the accuracy of the result? — Very dissatisfied to very satisfied.
8. How important is response speed when choosing a service? — Not at all important to extremely important.
9. How difficult was it to find the setting? — Not at all difficult to extremely difficult.
10. I would choose this workflow again. — Strongly disagree to strongly agree.
11. The follow-up reflected the answers I provided. — Strongly disagree to strongly agree.
12. I understood why I received this outcome. — Strongly disagree to strongly agree.

Each pattern contains one idea. If an item says a service was “fast and accurate,” split it so respondents can evaluate speed and accuracy independently.

Reusable coding template

These codes are a reproducible example, not a required standard. Record the displayed order, stored value, status, denominator rule and direction in the data dictionary. Preserve original responses when recoding an item.

Worked item revision

Draft: “The checkout was fast and easy.” A respondent who found it fast but confusing cannot answer precisely. Split it into “The checkout was fast” and “The checkout was easy to understand,” use the same labelled agreement options, then pretest both with target respondents.

Analysis guardrails

• Report ordered counts and percentages before reducing a single item to one summary number.
• For a multi-item score, publish the combination rule, reverse-coding rule, missing-data rule and evidence that the items belong together.
• Do not treat numeric codes as proof that adjacent categories are equal distances apart.
• Revalidate comparisons when wording, labels, order, language or collection mode changes.

Transparent method

We compared current vendor guidance with an institutional catalogue of response labels, retained points that agreed with basic question-design practice, and converted them into a decision matrix, worked revision and explicit coding example. We did not adopt advice to switch scale direction mid-survey because inconsistent polarity can create avoidable response errors.

Limitations

This guide does not validate a psychological construct, prove measurement invariance or prescribe one universal point count. Sensitive topics, accessibility needs, language and culture can change how categories function. Cognitive interviews and a pilot remain necessary for consequential measures.