What is a Likert scale, and how to interpret the results

If you've ever answered "Strongly agree / Agree / Neither agree nor disagree / Disagree / Strongly disagree" on a survey, you've already used a Likert scale. It's probably the most common question format in opinion, workplace climate, and satisfaction surveys, and it still gets built and interpreted incorrectly fairly often.

What a Likert scale actually is

The Likert scale (named after psychologist Rensis Likert, who proposed it in 1932) measures how much someone agrees or disagrees with a statement, using a symmetric range of options that goes from a negative extreme to a positive one, with a neutral midpoint. It doesn't measure a fact, it measures an attitude: how much someone agrees with "The checkout process was easy" is a Likert scale; "Did you buy something in the last 30 days?" is not, because there's no possible degree of agreement.

How many points the scale should have

The classic version has 5 points, though 7 points is also common when you need more granularity to detect subtle differences between groups. What almost never works well is using an even number of options (4 or 6): that removes the neutral midpoint and forces people to lean one way even when they genuinely don't have a formed opinion, which distorts the result instead of improving it.

• 5 points: enough for most cases, and faster to answer.
• 7 points: better when you need to compare groups with small differences between them.
• Avoid even-numbered scales (4, 6, 8 points): they force a position and hide genuinely neutral respondents.

How to interpret results without fooling yourself

The most common mistake is treating a Likert scale as if it were a pure numeric scale and only looking at the average. An average of 3.5 out of 5 could come from a group where everyone answered "3", or from a group evenly split between "1" and "5", two completely different situations that call for different responses. Always look at the full distribution of answers, not just the average: grouping into "agree" (4 and 5), "neutral" (3), and "disagree" (1 and 2) usually tells the real story better than a single number.

Ordinal, not interval (and why it matters)

Technically, a Likert scale is an ordinal scale: you know "agree" is more positive than "neutral", but you can't assume the distance between "agree" and "strongly agree" is mathematically equal to the distance between "disagree" and "neutral". In practice, for simple analysis (averages, group comparisons) almost no one strictly respects this distinction, and that's fine. But if you're going to run more formal statistical tests, using the median and non-parametric tests is more correct than treating the numbers as if they were centimeters on a ruler.

One item isn't enough to measure a complex concept

A single Likert question measures a single specific statement. If what you want to measure is something broader, like "commitment to the company" or "perceived service quality", a single question is usually insufficient and noisy. Standard practice is to use several related Likert items and combine them into an index, instead of resting the entire conclusion on one question.