In a survey questionnaire, the seemingly minor decision about the number of response options can affect data comparability, ease of completion, and how results are interpreted. A Likert scale requires not only selecting the number of points, but also precisely defining the categories and planning the analysis before the study begins.
How many points should a Likert scale have?
A Likert-type item is used to measure the intensity of an attitude, opinion, evaluation, or stated level of agreement with a statement. Rather than simply choosing “yes” or “no,” the respondent selects a position on an ordered continuum, for example from “strongly disagree” to “strongly agree.” In market research practice, five-point, seven-point, and four-point scales are used most often.
It is worth distinguishing between a single scaled question and a full Likert scale. A single question may read: “The brand provides information that is easy for customers to understand.” Only a set of several statements measuring the same construct, such as clarity of communication, trust, or perceived service quality, constitutes a scale in the methodological sense. This distinction matters during analysis because the results of several items can only be combined when they genuinely measure the same phenomenon.
A five-point scale is usually a good choice when the questionnaire needs to be simple and quick to complete. It allows respondents to express the direction of their opinion, indicate its strength, and select a neutral response. The response options may then be as follows: “strongly disagree,” “somewhat disagree,” “neither agree nor disagree,” “somewhat agree,” and “strongly agree.” The categories are clear, symmetrical, and easy to code.
A seven-point scale provides greater granularity. It is sometimes used when a study concerns products, brands, or experiences familiar to respondents, and participants can distinguish a moderate rating from a very strong one. However, additional points should not be created by mechanically adding unclear descriptions. The categories should be psychologically distinguishable to the respondent. If a respondent does not understand the difference between “fairly good,” “good,” and “very good,” a greater number of points does not improve measurement quality.
A four-point scale does not include a midpoint. It is used when the objective is to obtain a directional declaration and a neutral response has no interpretive value or may be used to avoid taking a position. However, the result of such a scale should be interpreted cautiously. Forcing a choice does not mean that the respondent genuinely has a formed opinion. Some people will select a positive or negative category at random in order to move on to the next question.
The choice of the number of points should reflect the nature of the issue being measured, the respondent profile, and the survey mode. A different design will work in a short satisfaction survey after contact with a helpline than in a study among B2B specialists who regularly use the solution being assessed. In both cases, the key consideration is what the evaluated object represents to the respondent and whether they have sufficient experience to differentiate their responses reliably.
When selecting the number of points, it is worth checking several conditions:
- whether the respondent is familiar with the product, service, brand, or experience being evaluated;
- whether the differences between adjacent categories are linguistically understandable;
- whether a midpoint is needed to describe neutrality or the absence of a directional opinion;
- whether the results will be compared with an earlier wave of the study;
- whether a consistent approach to response coding is used in comparable questions.
There is no single number of points that is appropriate for every project. In practice, a simple and unambiguous five-point scale is better than an elaborate seven-point scale that respondents cannot use correctly. Conversely, in a study where subtle differentiation of ratings is expected, an overly short set of response options may lead many declarations to cluster in the same categories.
How should categories be described and a Likert scale prepared for analysis?
A well-designed Likert-type item begins with a single statement measuring a specific aspect of an experience or attitude. The statement should concern one issue only. The question “The service is fast and competent” combines two distinct dimensions: response time and the substantive quality of the interaction. A respondent may rate the service as fast but incompetent, or vice versa. In this situation, the response does not indicate exactly what the rating concerns.
A better example of a Likert-type item for a customer experience study is the statement: “The information provided by the consultant was easy to understand.” This statement can be paired with five response options, from “strongly disagree” to “strongly agree.” A second statement may measure a separate area: “The consultant proposed a solution suited to the case.” Closed-ended questions designed in this way later allow the clarity of communication and the suitability of the solution to be analyzed separately.
Category descriptions must be symmetrical. If the negative end uses the wording “strongly disagree,” the positive end should use “strongly agree.” An asymmetrical set of response options may unintentionally suggest a particular evaluation. Problems also arise when middle categories are unclear, for example “average,” “hard to say,” or “no opinion,” even though they are intended to mean different things.
In market research, neutrality should be separated from the inability to provide an evaluation. A neutral response usually means a position in the middle of the scale: the respondent evaluates the phenomenon neither positively nor negatively. “Don’t know” or “not applicable,” on the other hand, indicates a lack of information on the respondent’s side. A person who has not used a given service should not be forced to select the midpoint. Such a response should be placed outside the main scale and treated analytically as no evaluation rather than as a neutral opinion.
Response coding should be determined before the study is launched. On a five-point agreement scale, consecutive values from 1 to 5 are most commonly assigned, with a higher value indicating stronger agreement. However, this is not a technical requirement but a principle for organizing the data. The most important thing is for the coding direction to be identical across questions that form the same index, unless a reverse-worded item is used. A reverse-worded item, for example “Obtaining information required considerable effort on my part,” requires subsequent recoding before a combined score is calculated.
In repeated studies, consistency in scale design throughout the study and across successive measurement waves is more important than the number of points itself. Changing from five to seven categories, changing endpoint descriptions, or reversing the scale direction makes it more difficult to compare results over time. If measurement is conducted on a recurring basis, the scale design should remain stable unless the change is intentional, documented, and taken into account in interpretation.
Preparing a Likert scale for analysis can be organized into the following steps:
- Define the single dimension that a given statement is intended to measure, such as trust, ease of use, or quality of interaction.
- Write a short statement in language understandable to the target group.
- Select the number of points and category descriptions that form a logical and symmetrical sequence.
- Add a “don’t know” or “not applicable” response if some respondents may not have a basis for evaluation.
- Establish coding rules and rules for recoding items worded in the opposite direction.
- Test the questions with a small group of people similar to the actual respondents.
Testing is particularly important when different measurement scales appear in the same instrument. A respondent may use an agreement scale efficiently but misinterpret a frequency or importance scale. It is also worth limiting changes in response direction across consecutive questions. When the left side indicates a negative rating in one question and a positive rating in the next block, the risk of automatic responses increases.
Analysis begins with data quality checks. The number of missing responses, the share of “don’t know” responses, the distribution of responses at each point, and whether respondents mechanically selected the same category across a long series of questions should all be examined. A high proportion of midpoint responses is not in itself an error. It may indicate genuine ambivalence, a lack of strong opinions, an unclear statement, or insufficient experience with the object being assessed. Determining the reason requires consideration of the question wording and the research context, rather than automatic data correction.
How can Likert scale responses be analyzed without misinterpretation?
Responses to a single Likert-type item are ordinal. This means it is known which response expresses a more positive or more negative evaluation, but it is not certain whether the distance between consecutive categories is identical for all respondents. Moving from “somewhat disagree” to “neither agree nor disagree” does not necessarily represent psychologically the same degree of change as moving from “somewhat agree” to “strongly agree.”
For this reason, the basic form of presenting results is the response distribution. It shows how many people selected each category and makes it possible to see the concentration of ratings, distribution asymmetry, and the share of neutral responses. For many questions, combining categories is also useful. Positive responses, meaning the two highest points on the scale, can be compared with negative responses, meaning the two lowest points, while retaining the midpoint as a separate category.
This method of presentation only makes sense when the aggregation rules are consistent. If positive responses in one question include the categories “somewhat agree” and “strongly agree,” the same definition should be used when comparing other statements measured on an identical scale. The midpoint should not be combined with positive ratings merely because doing so improves the appearance of the result. Neutrality is not agreement.
In practice, the arithmetic mean of responses to a single question is often used because it facilitates comparison across multiple items. However, it is important to remember how such a mean differs from a full description of the distribution. Two groups may obtain similar means even though one is dominated by midpoint responses and the other shows a clear divide between very positive and very negative ratings. For this reason, the mean alone should not replace a distribution table or a chart showing the response structure.
The following indicators are particularly useful in Likert scale analysis:
- the share of responses in each category, in order to see the full structure of declarations;
- the median, which indicates the middle position in an ordered set of responses;
- the mode, meaning the most frequently selected category;
- the share of positive and negative ratings, provided that the category combination was defined before the analysis;
- the mean, used cautiously as a synthetic comparative indicator;
- the percentage of “don’t know” or “not applicable” responses, which indicates whether respondents are able to evaluate a given issue.
If several questions are intended to form a single index, their substantive consistency should first be checked. Ratings concerning different phenomena should not be summed solely because the same scale was used. Statements about ease of purchase, product quality, and trust in a brand may have the same response options but measure different constructs. Combining them into a single score obscures information rather than organizing it.
Reverse-worded questions require recoding. If a higher score is always intended to indicate a more positive evaluation, responses to a negatively worded item must be recalculated in the opposite direction. Omitting this step may lower the index score or lead to apparent contradictions between questions. Recoding should be described in the analytical documentation so that the result can be reproduced.
When comparing results between groups, differences should be interpreted in relation to sample sizes and response structures. A difference in means alone does not show whether the change applies to a broad group of respondents or results from a small number of extreme ratings. In B2B research, this is particularly important when the segments being analyzed are small and a single response can clearly affect the overall result.
What most often distorts results when a Likert scale is used?
The most common error is imprecise statement wording. A Likert scale will not fix a question that contains ambiguous concepts, combines two issues, or assumes knowledge the respondent does not have. The wording “The offer is attractive” may refer to price, quality, availability, terms of cooperation, or brand image. The responses will be difficult to compare because survey participants may be evaluating different elements.
A second pitfall is confusing a neutral response with a lack of opinion. A midpoint such as “neither good nor bad” should not replace categories such as “I have not used it” or “hard to say.” When a person without experience selects the middle of the scale, the result suggests a neutral evaluation, even though it actually indicates no contact with the area being studied. This makes interpretation more difficult and may inflate the share of declared neutrality.
Excessive use of reverse-worded questions can also be problematic. Their purpose is sometimes to limit automatic response selection, but in practice, complex negations increase the risk of misunderstanding. The sentence “I do not believe that the process was not intuitive” requires additional effort from the respondent and can easily lead to an error. If a reverse-worded item is needed, it should be phrased simply and without double negatives.
A Likert scale can also produce acquiescence bias. Some respondents tend to agree with statements regardless of their content, particularly when questions concern socially desirable behavior, relationships with a brand, or their own competencies. This issue can be mitigated by using clear, balanced statements and avoiding leading language. However, many negative formulations should not be introduced solely to formally balance the questionnaire.
It is also worth distinguishing an agreement scale from other measurement tools. The question “How often do you use the service?” requires a frequency scale, while the question “How important is this element?” requires an importance scale. Using the same response format for different phenomena may look visually consistent, but it weakens the fit between the question and the construct being measured. Closed-ended questions are valuable when their response options correspond precisely to what the wording asks about.
A further limitation is comparing results across studies that use different scales. A result on a five-point scale is not automatically comparable with a result on a ten-point scale, even if both concern the same issue. The number of available categories, the way they are described, and the presence or absence of a midpoint all differ. For historical comparisons, the same question design, order, and category descriptions should be maintained.
Frequently asked questions
How many points should a Likert scale have?
A Likert scale most often has five or seven points, and the choice depends on how precisely respondents can differentiate their ratings. A five-point scale is usually easier to understand, while a seven-point scale may provide a more detailed picture of opinions. More important than the number of points itself are clear category descriptions and maintaining the same scale throughout the study.
Should a Likert scale have a midpoint?
A midpoint is worth using when a neutral opinion is a possible and important response. It should not replace a “don’t know” or “not applicable” option, because a lack of experience with the object being evaluated is not neutrality. A Likert scale without a midpoint can be used deliberately, but it forces respondents to declare the direction of their opinion.
How should Likert scale responses be analyzed?
Analysis should begin with the distribution of responses for each category, followed by an examination of the share of positive, neutral, and negative ratings. A Likert scale can also be described using the median, mode, and, cautiously, the mean, but a single indicator should not replace the full distribution. When combining several questions into an index, it is necessary to verify that they all measure the same dimension and that reverse-worded items have been recoded correctly.
Ask about designing a questionnaire for your study. Hume’s Institute will help select a Likert scale, define the response options, and prepare data analysis rules before the study is conducted.