Likert Scale: Definition, Examples and How to Analyze It
- Use a Likert scale to turn agreement with a statement into ordered numbers you can count, compare and track.
- Choose 5 points for most customer surveys: agree-disagree scales with 5 categories gave higher-quality data than 7 or 11 in European Social Survey tests [Sociological Methods & Research, 2014].
- Put one idea in each statement, and keep the same labels and the same direction for every item.
- Report medians, category shares and a top-2-box score, because a single Likert item is ordinal data.
- Reverse-score negatively worded items before you sum or average them, or the combined score is wrong.
- Pair the scale with one open question, because a Likert score shows how much people agree, not why.
A Likert scale is a survey rating scale that asks people how strongly they agree or disagree with a statement, on an ordered set of labeled points, usually 5 or 7. Each label gets a number, so opinions become data you can count, compare and track. The psychologist Rensis Likert introduced the method in A Technique for the Measurement of Attitudes in 1932 [Archives of Psychology, 1932].
In eCommerce, Likert scales sit inside satisfaction surveys, post-purchase surveys and on-site polls, and they are one of the simplest ways to measure the eCommerce customer experience. This guide covers what a Likert scale is, how many points to use, Likert scale examples, how to write the questions and how to analyze the answers.
What is a Likert scale?
The difference matters when you analyze the results. A Likert item is defined as one statement plus its response options, for example "The checkout was easy to use", rated from Strongly disagree to Strongly agree. A Likert scale, in Likert's original sense, combines several items that measure the same idea, such as four statements about delivery, into one score. A combined score is more stable than any single answer.
The standard 5-point version codes its labels like this:
- 1: Strongly disagree
- 2: Disagree
- 3: Neither agree nor disagree
- 4: Agree
- 5: Strongly agree
The answers are ordinal data. They have a clear order, but nobody can prove that the gap from Agree to Strongly agree equals the gap from Neutral to Agree. That one fact decides which statistics you can use, as the analysis section shows.
5-point vs 7-point Likert scale: how many points should you use?
More points do not mean better data. Revilla, Saris and Krosnick compared agree-disagree scales of 5, 7 and 11 categories in four multitrait-multimethod experiments and concluded that researchers should offer 5, because the longer versions yielded lower-quality data [Sociological Methods & Research, 2014]. The table compares the common formats.
| Format | Example labels | Best for | Watch out for |
|---|---|---|---|
| 4-point (forced choice) | Strongly disagree to Strongly agree, no midpoint | Questions where you need a lean, one way or the other | People with no opinion pick a side at random |
| 5-point | Strongly disagree, Disagree, Neutral, Agree, Strongly agree | Most customer and on-site surveys, mobile | Less room for fine shades of opinion |
| 7-point | Adds Slightly disagree and Slightly agree | Academic research, expert or engaged audiences | Labels get harder to tell apart on small screens |
| Unipolar | Not at all satisfied to Extremely satisfied | Measuring how much of one quality is present | No negative side, so it cannot show active dislike |
| Bipolar | Very dissatisfied to Very satisfied, neutral in the middle | Measuring direction and strength of a feeling | The midpoint mixes "neutral" with "no opinion" |
For the wider set of rating, ranking and comparative formats, see how to choose a survey scale.
Likert scale examples and question types
| Response type | 5-point labels (lowest to highest) | eCommerce example |
|---|---|---|
| Agreement | Strongly disagree to Strongly agree | "The product page answered my questions." |
| Satisfaction | Very dissatisfied to Very satisfied | "How satisfied are you with your delivery?" |
| Frequency | Never to Always | "How often do you use the wishlist?" |
| Importance | Not at all important to Extremely important | "How important are free returns when you choose a store?" |
| Likelihood | Very unlikely to Very likely | "How likely are you to buy from us again?" |
| Quality | Very poor to Excellent | "How would you rate the packaging?" |
Online stores use these items most in customer satisfaction surveys, where a satisfaction item feeds the customer satisfaction score (CSAT). The same format measures employee engagement, product feedback and usability. The Net Promoter Score looks similar, but it uses a 0 to 10 likelihood scale with its own formula.
How do you write good Likert scale questions?
-
Put one idea in each statement"Delivery was fast and cheap" mixes two ideas, so a "Disagree" tells you nothing. Split it into "Delivery was fast" and "Delivery was good value".
-
Use plain, neutral wordsAvoid jargon and leading adjectives. "Our amazing support team helped you" pushes people toward Agree; "Support solved my problem" does not.
-
Keep the scale identicalUse the same number of points, the same labels and the same direction for every item. If 1 means Strongly disagree in one question, it means that everywhere.
-
Avoid absolute words"The site never loads slowly" is hard to agree with honestly. "The site loads quickly" lets people rate what they actually saw.
-
Reverse a few items, and record whichSome negatively worded statements stop people from clicking Agree on every row. Mark them in your codebook before the survey goes live.
That last step is where many teams go wrong. They add negatively worded items such as "The checkout was confusing" to catch automatic agreement, then forget to reverse-score them before they average the scale. A "Strongly agree" to the confusing-checkout item then counts as 5, the same as a "Strongly agree" to "The checkout was easy". The combined score drifts toward the middle, the reliability check fails, and the team reports that customers feel lukewarm when the real answers were split. Recode reversed items (6 minus the score on a 5-point scale) before any total.
How to analyze Likert scale data
-
Code the labelsGive each label a number, starting at 1 for the most negative, and recode reversed items.
-
Describe the distributionReport the median, the mode and the percentage of answers in each category. A mean hides a split between lovers and haters.
-
Calculate a top-2-box scoreAdd the two most positive categories and divide by all answers. It is one number a stakeholder can track month to month.
-
Visualize itA diverging stacked bar chart puts disagreement left and agreement right, so you see the balance at a glance.
-
Test the differencesCompare two groups with Mann-Whitney U, three or more with Kruskal-Wallis, and relationships with Spearman correlation. For summed scale scores, analysis of variance (ANOVA) or regression analysis are common, with ordinal logistic regression for single items.
Top-2-box score is defined as the share of respondents who pick one of the two most positive points on the scale. The formula is: Top-2-box score = (Agree + Strongly agree) ÷ total responses × 100. Worked example: 400 shoppers rate "My order arrived when promised". The answers are 20 Strongly disagree, 40 Disagree, 60 Neutral, 160 Agree and 120 Strongly agree. Top-2-box = (160 + 120) ÷ 400 × 100 = 70%. The bottom-2-box share is 60 ÷ 400 = 15%, and the median answer is Agree.
GetMaineLobster, a US online seafood retailer, used Omniconvert surveys to ask leaving visitors what held them back, then answered those objections with on-site personalizations. Its conversion rate rose 51%, exit visitors generated an additional 7% of revenue, and email leads grew 9% [Omniconvert, GetMaineLobster case study]. The case study describes an exit survey, not a Likert battery, but the lesson carries over: a score pays only when the weakest answer becomes a change you test.
What are the advantages and limitations of a Likert scale?
The advantages are practical: the format is familiar, the coded answers fit any spreadsheet, and the same items run in email, on-site and in-app surveys, so you can compare segments and track trends.
The limits come from people. Acquiescence bias makes some respondents agree with whatever they read, social desirability bias pulls answers toward what sounds good, and central tendency bias keeps others away from the ends of the scale. Reversed items, anonymity and short surveys reduce all three; see response bias for the full list. To learn why a score is low, add one open question and read the answers, or group them with natural language processing (NLP) for survey results. That pairing is how you find customer pain points instead of only measuring them.
Likert scale FAQs
A Likert scale is a set of statements that people rate on the same ordered scale, usually from Strongly disagree to Strongly agree. Each answer gets a number, so you can count, compare and track opinions over time. The psychologist Rensis Likert introduced the method in 1932 as a simpler way to measure attitudes.
A single Likert item is ordinal: the answers have an order, but nobody can prove the gaps between them are equal. Report medians, modes and response shares for single items. When you sum or average several related items into one scale score, many researchers treat the result as approximately interval and use t-tests or ANOVA, provided they state that assumption.
Code each label as a number, report the median and the share of answers in each category, and calculate a top-2-box score: Agree plus Strongly agree as a share of all answers. Show the distribution in a diverging stacked bar chart. To compare groups, use non-parametric tests such as Mann-Whitney U or Kruskal-Wallis for single items, and keep t-tests or ANOVA for summed multi-item scores.
Include a midpoint when a neutral opinion is a real answer, which it is for most customer questions. Remove it, with a 4-point or 6-point scale, only when you need respondents to lean one way. A forced choice cuts fence-sitting, but people with no opinion then pick a side at random, which adds noise you cannot see in the results.
A 7-point scale adds two in-between options, so it captures finer shades of opinion and suits engaged or expert audiences. A 5-point scale is faster to answer and easier to read on a phone. For agree-disagree questions, experiments in the European Social Survey found that 5 categories gave higher-quality data than 7 or 11 [Sociological Methods & Research, 2014].
No. The Net Promoter Score asks how likely a customer is to recommend you on a 0 to 10 scale, so it is a likelihood rating scale, not an agree-disagree Likert item. Both produce ordinal data and both fit in the same survey, but NPS has its own formula: the percentage of promoters minus the percentage of detractors.
Before you send your next survey, cut each statement down to one idea, fix the scale at 5 labeled points, and write down which items are reversed. When the answers come in, report the median and the top-2-box share for each statement, then sort the statements from lowest to highest score.
The statement at the bottom of that list is your next hypothesis. Add one open question beside it to learn why customers disagree, then test a change that answers them. Omniconvert Explore runs the survey and the A/B test in one place, so the score and the fix live side by side.
Turn Likert scores into tested changes
A low score tells you where customers disagree. Omniconvert Explore runs on-site surveys, A/B tests and personalization in one tool, so you can ask, find the weakest statement, and test the change that fixes it.