Guide · Questionnaire design
Likert scale: how to write, build and analyse one
A Likert scale measures attitudes: you present statements and respondents say how much they agree – usually on five or seven points from “strongly disagree” to “strongly agree”. This guide shows how to write clear statements, pick the right number of points, deal with the midpoint and “no answer”, analyse the results and build a Likert grid in 121surveys.
Building and testing is free. To publish, you choose a paid plan.
Understanding and using the Likert scale
What is a Likert scale?
The Likert scale goes back to the social psychologist Rensis Likert, who introduced it in 1932. It consists of several statements on the same topic, all rated with the same response points. Strictly speaking, only the set of statements is the scale; a single statement with five points is more precisely called a Likert item. Typical uses are employee surveys, customer satisfaction, course evaluation and product tests – anywhere you want to measure an attitude rather than just a yes or no.
Writing the statements
Each statement covers exactly one idea. “The software is fast and easy to navigate” mixes two things – anyone who likes only one of them cannot answer honestly. Keep it short, concrete and in your respondents' language, with no double negatives and no loaded wording that suggests an answer. Add a few reverse-worded statements (“I often spend a long time looking for features”) so that mindless clicking stands out – and reverse their values in the analysis before you combine statements.
5 or 7 points?
Five points are the classic choice: quick to answer, easy to read on a phone and fine-grained enough for most topics. Seven points separate more finely when your respondents know the subject well and small differences matter. More points do not automatically make an answer more precise – they take time, and the boundaries between points blur. What matters is that every statement in a scale uses the same points and that each point is labelled, not just the two ends.
The midpoint and “no answer”
An odd number of points has a midpoint (“neither agree nor disagree”). It is honest for people without a clear opinion, but it is also chosen to avoid committing. An even number forces a direction – useful when a tendency matters more to you than a neutral camp. Either way, you need a separate “no answer” option for people who cannot judge a statement. Placed inside the scale it distorts averages, so it belongs off to the side and counts towards no figure.
Analysing Likert data
Start with the frequency distribution for each statement: how many people chose which point? For a quick comparison, use the top-2 box – the share of the two agreeing points – and, as its mirror image, the bottom-2 box. Averages are common, but strictly speaking Likert responses are ordinal: the gap between “agree” and “strongly agree” need not be the same as between two other points. So always report the distribution or the median alongside the mean, and compare means only between statements on the same scale.
Common mistakes
Frequent problems are too many statements in one grid (attention drops, straight-lining rises), mixed scales within one block, unlabelled points and statements that only experts understand. Just as risky: statements that all point in the same direction, and a fixed order in which the first statement always gets the most attention. Test your questionnaire beforehand with a few people from your target group and watch where they hesitate or ask questions.
How to build a Likert scale in 121surveys
Rating grid for statements
Statements are the rows, points are the columns – each point with its own label. For a single statement, use the rating scale with labelled ends.
Shuffle the order per person
If you like, the rows of a grid appear in a random order for each person, so that no statement benefits from being first.
“No answer” on every row
One switch adds a “prefer not to answer” option to every statement. It sits next to the scale and is left out of every average.
Analysis and data quality
For each statement you see the distribution and the mean, compare groups in a crosstab, and the quality check flags people who click straight down a grid.
Frequently asked questions
Is a Likert scale ordinal or interval?
Strictly speaking, ordinal: the points have an order, but the gaps between them are not guaranteed to be equal. In practice, means are still often calculated, especially when several statements are combined into an index. When you do this, always show the distribution too, so nobody reads more into an average than it contains.
How many statements should a Likert scale have?
As many as the topic needs – and as few as possible. Several statements about the same trait make the measurement more stable than a single one. Long grids are tiring, though; it is better to spread many statements over several pages or topic blocks.
Which labels should I use for the points?
For agreement, for example: strongly disagree, disagree, neither agree nor disagree, agree, strongly agree. The same principle applies to frequency (never to always) or satisfaction (very dissatisfied to very satisfied): symmetrical, even steps, every point written out.
How does it differ from a semantic differential?
With a Likert scale, respondents rate statements by how much they agree. With a semantic differential, they place an object between two opposite qualities, such as “modern – old-fashioned”. The differential tends to measure image and perception, the Likert scale attitudes and opinions.
Can I try Likert scales in 121surveys for free?
Yes. You build the questionnaire and try it out via the test link before you publish anything. Test responses are left out of every statistic, so your later results stay clean.
Your next Likert scale, built properly
Set up your statements in a rating grid, shuffle their order and try everything for free via the test link.
