A Likert scale asks people to choose one point on an ordered set of answer options, such as “strongly agree” through “strongly disagree.” It is one of the most common ways to measure attitudes and opinions in market research, because it turns a judgment into a value you can compare across groups. This guide covers example questions, how to set up the scale, how to write balanced items, and how to analyze the results without claiming more than they show.
What a Likert scale is
Rensis Likert introduced the approach in 1932. A Likert scale presents a statement or question with a set of labeled, ordered answer options. Most scales have five or seven options, though the format can range from about four to nine. The options describe a single attribute, such as agreement, satisfaction, likelihood, frequency, or importance.
Example Likert scale questions
Each example below pairs a question with its answer labels. Keep the labels for one question consistent with the others in the same survey.
| Type | Example question | Answer labels (5-point) |
|---|---|---|
| Agreement | “The product is easy to set up.” | Strongly disagree, Disagree, Neither agree nor disagree, Agree, Strongly agree |
| Satisfaction | “How satisfied are you with the delivery time?” | Very dissatisfied, Dissatisfied, Neutral, Satisfied, Very satisfied |
| Likelihood | “How likely are you to buy this product in the next three months?” | Very unlikely, Unlikely, Neutral, Likely, Very likely |
| Frequency | “How often do you compare prices before buying [category]?” | Never, Rarely, Sometimes, Often, Always |
| Importance | “How important is delivery speed when you choose a supplier?” | Not at all important, Slightly important, Moderately important, Very important, Extremely important |
Odd or even: choosing the number of points
An odd scale, such as 3, 5, or 7 points, includes a neutral midpoint. An even scale, such as 4 or 6 points, forces respondents to lean one way.
- Five points is the most common choice. It balances ease of answering with enough detail to be useful.
- Seven points captures finer differences, but takes more effort, which can matter on a phone.
- Three points is quick and easy, but gives little detail.
- Even scales are useful when a neutral answer would hide a real opinion. They can frustrate people who genuinely have no view, so use them with care.
Pick one number of points for a concept and keep it. Changing from five points to seven partway through a survey makes results hard to compare.
Unipolar and bipolar scales
A bipolar scale runs between two opposite ends, such as “strongly disagree” and “strongly agree,” with a neutral point in the middle. Use it when you want to see both sides of an opinion.
A unipolar scale measures how much of something a person has, from none to a lot. For example: “How appealing is this design to you?” with options from “Not at all appealing” to “Extremely appealing.” Use it when the question is about intensity rather than direction.
Writing balanced Likert items
Clear items produce clearer answers. These rules apply to most Likert questions:
- Ask about one idea per item. Split “satisfied with the features and price” into two items.
- Use plain words and specific timeframes. “In the last week” is more precise than “usually.”
- Label every option, and keep the options balanced. Include the same number of positive and negative options, and make sure they do not overlap.
- Mix positive and negative statements. If every statement points the same way, some people will agree with all of them without thinking. Reverse-code the negative items during analysis.
- Choose statements or questions with care. Classic Likert items are statements (“I am satisfied with the quality of this product.”). A question often gets a clearer answer and less automatic agreement: “How satisfied are you with the quality of this product?” Using both in one survey works well, but do not rely on statements alone.
The guide to leading and biased survey questions covers how to remove wording that pushes people toward an answer.
Common response biases
Scales can produce predictable patterns that are not about the topic itself:
- Acquiescence. Some people agree with almost everything. Mixing positive and negative items helps reveal this.
- Midpoint overuse. Some people choose the neutral option when they are unsure, tired, or uninterested. A high share of neutral answers can be a warning sign, not a finding.
- Social desirability. People may pick the answer that makes them look good, especially on sensitive topics.
- Fatigue. Long grids of similar items lead to straight-lining, where the same answer is picked for every row.
Analyzing Likert results
Likert answers are ordinal. The order of the options matters, but the distance between “agree” and “strongly agree” is not known to be the same as the distance between “neutral” and “agree.” That shapes which summaries are fair.
Single items: distributions, median, and mode
For a single item, start with the distribution: the share of people who chose each option, shown as a bar chart. The median and mode are defensible summaries. You can also report the share who chose the two most positive options, often called “top-two-box,” as long as you say which options you included. This is how a CSAT score is usually reported.
Averaging a single item is common, but it treats the scale as if the gaps were equal. If you do report a mean for one item, state that assumption and show the distribution alongside it.
Composite scores: means and reliability
When several related items measure one idea, you can combine them into a composite score, such as the average of four agreement items about trust. Reverse-code any negative items first, so that higher always means more of the idea. Check whether the items hang together, for example with Cronbach’s alpha. The pilot testing guide explains why alpha from a small pilot is unstable and should be read with caution.
Treating the composite as an interval measure, and reporting a mean and standard deviation, is a common choice. State that assumption in your report.
Tests for ordinal and interval data
For ordinal data, use tests designed for ranked or categorical results, such as Spearman’s correlation or a chi-square test of independence. For composite scores treated as interval data, Pearson’s correlation or analysis of variance (ANOVA) are common. Check your software’s documentation for the exact options, and report which test you used and why.
Before you compare two groups, check how many people are in each. A small difference between two small groups is often just noise.
Common mistakes
- Mixing scale lengths for one concept. A five-point question followed by a seven-point question on the same topic is hard to compare.
- Leaving endpoints unlabeled. Respondents need to know what the ends mean.
- Overlapping options. “Somewhat” and “a little” can mean the same thing to a respondent.
- Reporting averages of single items without the distribution. The same average can come from very different patterns.
- Treating a midpoint as a finding. A large share of “neutral” answers often signals unclear wording or disengagement.
For the broader design of a questionnaire, including how scale questions fit with the rest, see the market research survey questions guide.