What are Likert Scales?

Likert Scales are commonly used to understand psychometric traits, attitudes, beliefs, perceptions, and actions of the survey takers. Participants are given a scale to measure these attributes and asked to rate the intensity in which fall along the scale. If you’ve ever answered a question on a scale of extremely disagree → extremely agree, you’ve used a Likert scale before!

Likert scales allow you to collect data on a granular level so that you can splice and analyze these responses to a greater degree. Sometimes, a simple “yes” or “no” answer doesn’t give you enough insights. You may need to know the level of degree of the attribute in order to take actionable next steps.

Picture this use case: you own a D2C clothing line and want to hire an influencer to promote your brand. You’re on a limited budget and need to make sure you select the right influencer. If the influencer doesn’t fit within your brand image, will it be a waste of your money? Do your customers even care? Using Likert scales, you can gauge the public sentiment on influencers and gather data to help you make decisions.

These are just two examples that can be used in this situation.

Question: 
- How much do you agree or disagree with the following statement
  - "I pay attention when influencers promote fashion brands."

Options:
- Strongly disagree
- Somewhat disagree 
- Neither agree or disagree 
- Somewhat agree 
- Strongly agree

Question: 
- What is your perception of (Influencer’s Name)?

Options:
- Very negative 
- Somewhat negative 
- Neutral / I don’t know who this is
- Somewhat positive
- Very positive

What type of scale do I use?

Likert scales can be presented in two ways: using a bipolar or unipolar scale.

A bipolar scale gives you the advantage of measuring intensity from both directions of an attribute, usually with a neutral option in the middle. This allows the participant to rate either side of the attribute and measure the intensity of the direction they choose. For example, you will receive data on the level the participant agrees vs disagrees on a question or statement.

A unipolar scale gives you the advantage of measuring the intensity in one direction of an attribute. This allows the participant to rate the attribute on a more granular level. For example, you will receive data on the level at which they agree on a question or statement.

There are pros and cons to using both types of scales, and differences of opinions on which is best. Two examples of the debates for both types of scales: with bipolar scales, some people don’t like the neutral option because it allows users to lazily answer, making the data useless. With unipolar scales, some people think the difference levels are so minute that it makes it hard for participants to select an answer, making the data useless. You must decide which scale is more appropriate for your use case so that you can receive actionable insights.

Taking one of the influencer questions from above, the difference using the two scales would look as such as:

Question:
How much do you agree with the following statement: I pay attention when influencers promote fashion brands.

Bipolar Scale

Options:
- Strongly disagree 
- Somewhat disagree 
- Neither agree or disagree 
- Somewhat agree 
- Strongly agree

Unipolar Scale

Options:
- Not at all agree
- Slightly agree 
- Somewhat agree
- Extremely agree

Choosing the number of points for your scale

Once you select which type of scale you’re going to use, the next step is choosing the number of points for the scale. Likert scales usually range from a 3-10 point scale, 4, 5 and 7 being the most common points.

Once again, there is a heavy debate on the most appropriate number of points for a scale. Typically, bi-polar scales are an odd number because they contain a neutral option with an even amount measuring each direction. Unipolar scales can be any number, as it's up to you to choose the level of granularity to measure.

We lay out the pros and cons for including more or less points on your scale below. Use it as a guide to select the amount that feels appropriate for your needs. More importantly: once you select a point scale to use, be consistent with it throughout the rest of the survey and future surveys; it’s harder to compare the responses from two questions if one uses a 4 point scale and the other uses a 7 point scale.

Less Points (i.e. 3-5 point scales)

More Points (i.e. 6-10 point scales)

Keep the measurement on your scale unbiased

No matter what point scale you select, it’s important to have consistent measurements across your scale. Inconsistency can confuse the respondent, bias responses, and ruin the data. Consistent measurements happen in two ways 1) between the intensity along the entire scale and 2) between the intensity between each point.

Intensity along the entire scale means the points you offer fairly measure the possibility of responses.

For example, on a bipolar scale, you should have 1 neutral option with the same amount of positive and negative options in either direction. On a 5-point scale this shows as follows:

Options:
- Strongly disagree 
- Somewhat disagree 
- Neither agree or disagree 
- Somewhat agree 
- Strongly agree

Here there are 2 negative responses, 1 neutral response, and 2 positive responses. This gives the respondent a fair chance to rate their positive and negative sentiment.

On a unipolar scale, you should have 1 option that equates to “0%” and one option that equates to “100%”. On a 4-point scale it shows as follows: On a 4-point scale it shows as follows:

Options:
- Not at all agree
- Slightly agree 
- Somewhat agree
- Extremely agree

Here “not at all agree” and “extremely agree” bookend the response options. This gives the respondent a fair chance to choose a rating between 0-100%.

Intensity between each point

means that the measurement between each point remains the same along the entirety of the scale. If the responses don’t represent the entire spectrum of options fairly, it can confuse the respondent and bias the results. The same should be true for both bipolar and unipolar scales. Using a bipolar scale in the examples below:

#1: In this example below, there isn’t a fair response for the entire spectrum of the scale. “

Options:
- Completely disagree 
- Strongly disagree 
- Neither agree or disagree 
- Strongly agree 
- Completely agree

Completely’ and ‘Strongly’ can be interpreted to mean the same thing, leaving no real middle ground between the ‘strongly’ --> ‘neither agree or disagree’ options on both sides of the scale. This example specifically biases the results in an extremely negative or positive reaction and leaves out a more lukewarm sentiment.

#2: In this example, there is poor consistency across the responses in the scale, leaving a lot of room for interpretation on the weight of each response.

Options:
- Extremely disagree 
- Slightly disagree 
- Neither agree or disagree 
- Strongly agree 
- Completely agree

There is a huge jump between ‘Extremely disagree’ and ‘Slightly disagree,’ but a small jump between ‘Strongly agree’ and ‘Completely agree.’ This makes both answering this question and analysing the results messy, because everyone who looks at the scale can interpret it in a different way. This example specifically biases respondents to give extremely positive results.

#3: In this example, there is a fair response for the entire spectrum of the scale AND there is no room for interpretation on the weight of each response.

Options:
- Extremely disagree 
- Somewhat disagree 
- Neither agree or disagree 
- Somewhat agree 
- Extremely agree

This makes it easy for respondents to answer and it allows for clean data analysis because we know there is a definitive value for each response.

How measurement can go totally wrong

I recently received an opinion survey with the response options as follows:

Options:
- Outrage
- Dismay
- Disappointment
- I’m not concerned about this
- No opinion

This question is offering a 5-point scale with 3 points that are negative, 1 that is “positive” and 1 that is neutral. I’m also not sure the weight and difference between the three negative points, and I'm sceptical there even is one for the positive sentiment.

By framing the responses like this, the organizer is blatantly trying to bias the results with an overwhelming negative reaction. If respondents didn’t have an initial opinion on the matter, seeing 3 negative responses could force them into having a negative opinion.

Rate multiple statements at once

To make it easier on yourself and your participants: if you’re using the same Likert scale in multiple questions, you can rank multiple questions or statements rated at once instead of individually.

Using the influencer example and a 5-point bipolar scale, this can look like such:

Question: 
- Rate your perception of the each of the following influencers:

When you have them rate multiple statements at once, try and keep the sentence structure the same across each item being rated. To test if your participants aren’t just running through and clicking responses without reading the options, you can choose to flip the phrasing of one of the items. For example, if you have a bunch of statements being rated and all the statements are phrased positively, randomly phrase one statement negatively.

You should also make sure that all of the statements being rated together are all relevant to the same topic; just because the same Likert scale is being used doesn’t mean that the questions should be grouped together. In this example, only influencer perception is being rated. It doesn’t make sense to also have them rate their experience on your website, even if you want to use the same scale.

< Take Me to Question Bank