The Likert scale — "strongly disagree" to "strongly agree," typically across 5 or 7 points — is the most-used question format in survey design, and one of the most commonly misused. Here's how to build one that actually produces comparable, trustworthy data, with worked examples of exactly what the labels should say.
What it's actually for
A Likert scale measures agreement or intensity with a statement, not a question. That distinction matters: "I feel my manager listens to my concerns" (statement, rate agreement) behaves differently than "Does your manager listen to your concerns?" (question, pick yes/no). The statement format is what lets you compare responses across many items on one consistent scale — which is the entire point of using Likert scales instead of one-off custom questions for each thing you're measuring.
It's named after psychologist Rensis Likert, who introduced the format in 1932 specifically to measure attitudes that don't have a single objectively correct answer — satisfaction, agreement, sentiment. If what you're measuring has a factual right answer, a Likert scale is the wrong tool; use multiple choice or yes/no instead.
A complete 5-point Likert scale example
Here's exactly what a fully labeled 5-point agreement scale looks like — this is the version to copy if you just need a working example:
- Strongly disagree
- Disagree
- Neither agree nor disagree
- Agree
- Strongly agree
Applied to a real statement — "I have the tools and resources I need to do my job well" — a respondent picks one of those five labels, not a bare number. That's the whole mechanic.
A complete 7-point Likert scale example
The 7-point version adds two intermediate steps between the ends and the middle:
- Strongly disagree
- Disagree
- Somewhat disagree
- Neither agree nor disagree
- Somewhat agree
- Agree
- Strongly agree
Same statement, same mechanic, finer resolution. Whether that extra resolution is worth using is the next question.
Choosing the number of points
5-point is the standard default. It's fast to answer, easy for respondents to reason about, and every point can be labeled with a phrase that reads naturally — which becomes progressively harder as you add more points.
7-point gives finer-grained data and is worth it when you expect responses to cluster — e.g., most people will "agree" with something, and you want to distinguish mild from strong agreement within that cluster, rather than losing that distinction in one bucket.
Avoid even-numbered scales (4 or 6 points) unless you deliberately want to force a lean one way or the other by removing the neutral midpoint. That's a legitimate design choice in some cases — see the FAQ below — but it should be deliberate, not accidental.
Here's a fuller breakdown of when each scale length makes sense:
| Points | Has a midpoint | Best for |
|---|---|---|
| 4-point | No (forced choice) | You specifically want to eliminate the "neutral" dumping ground and force a directional lean |
| 5-point | Yes | The general-purpose default — fast, easy to label fully, easy to analyze |
| 6-point | No (forced choice) | Same forced-choice logic as 4-point, with slightly finer resolution on each side |
| 7-point | Yes | You expect answers to cluster and want to distinguish intensity within that cluster |
| 9-point or 10-point | Rare | Academic/clinical instruments needing very fine granularity — usually more precision than a business survey needs or respondents can reliably provide |
| 11-point (0-10) | N/A — different format | Not really a Likert scale — this is NPS's format, standardized for a specific question, not a general-purpose agreement scale |
Beyond about 7 points, most respondents can no longer reliably tell you what makes an 8 meaningfully different from a 9 on a subjective feeling — the extra points add apparent precision without adding real signal.
Label every point, not just the ends
Labeling only "1 = Strongly disagree" and "5 = Strongly agree," with 2 through 4 left as bare numbers, forces every respondent to silently interpret the middle of the scale differently. Some will treat 3 as neutral, others as "leaning disagree." Label all five (or seven) points explicitly — using the exact wording in the examples above — and you remove that ambiguity entirely. It costs nothing and meaningfully improves consistency, both within one respondent's answers and across all your respondents.
Keep the scale direction consistent
If most of your statements are worded so that "agree" means positive sentiment, don't slip in one reverse-worded statement ("I do NOT feel supported by my team") without a clear reason. Reverse-wording is sometimes used deliberately to catch respondents who are straight-lining (clicking the same answer down the whole page without reading), but if you use it, flag it clearly in your own analysis — otherwise a reversed item will silently drag your averages in the wrong direction when you average it in with the rest.
Writing the statements themselves
- One idea per statement. "My manager gives me clear direction and recognizes my work" is two statements wearing one sentence — split it. If a respondent agrees with the first half and disagrees with the second, they have no way to answer accurately.
- Avoid absolutes. "My manager always listens" invites disagreement from anyone who can think of a single counterexample, which isn't what you're trying to measure. "My manager listens to my concerns" measures the same underlying sentiment without the trap.
- Keep it concrete. "I'm satisfied with my work environment" is vaguer than "I have the tools and resources I need to do my job well" — the concrete version tells you something you can act on.
- Write the statement, not a question. "I would recommend this product to a colleague" is a Likert statement. "Would you recommend this product?" is a yes/no question wearing a Likert scale's clothes — it'll frustrate respondents who feel strongly but not absolutely.
Example statements by use case
Employee engagement:
- "I understand how my work contributes to the team's goals."
- "I feel comfortable raising concerns with my manager."
- "I have the resources I need to do my job well."
Product feedback:
- "This feature was easy to find."
- "The product does what I expected it to do."
- "I would recommend this product to a colleague."
Customer satisfaction:
- "The support team resolved my issue quickly."
- "I felt the pricing was clearly explained."
- "Overall, this experience met my expectations."
Notice each one is a single, concrete, non-absolute statement a respondent can genuinely agree or disagree with — that's the pattern to replicate for your own statements.
Reading the results
Likert data is ordinal, not truly numeric — the gap between "agree" and "strongly agree" isn't guaranteed to be the same psychological distance as the gap between "neutral" and "agree." In practice, most teams average the numeric codes anyway (treating 1–5 as interval data) because it's a reasonable approximation at scale, but for small sample sizes, look at the distribution of responses, not just the mean. A cluster split evenly between "strongly agree" and "strongly disagree" can produce the same average as a survey where everyone answered "neutral," and those are very different situations that a single mean will hide from you.
For anything you plan to track over time or compare across teams, keep both the mean and the distribution (a simple bar count per label) in view — the mean tells you direction, the distribution tells you whether that direction is a real consensus or a split opinion.
Where it shows up in practice
Likert scales are the backbone of most employee engagement, culture, and satisfaction surveys, because they let you track the same statements over time and compare across teams or departments on one consistent scale. See our guide to anonymous employee feedback for how to pair Likert-scale design with the trust employees need to answer honestly, or our comparison of NPS vs. CSAT vs. CES for where a Likert-style agreement scale fits alongside those other metrics — Customer Effort Score is itself typically asked as a Likert agreement statement ("this was easy to do").
FAQ
What's a good example of a Likert scale statement? A concrete, single-idea statement written in the first person: "I have the tools and resources I need to do my job well," rated on a labeled scale from "Strongly disagree" to "Strongly agree." Avoid absolutes ("always," "never") and avoid bundling two ideas into one statement.
Should I include a neutral midpoint? Default to yes — an odd-numbered scale (5 or 7 points) with a "Neither agree nor disagree" midpoint. Some respondents genuinely are neutral or feel the statement doesn't apply to them, and removing the midpoint doesn't give them an opinion, it just forces a side that doesn't reflect how they feel. Use an even-numbered, forced-choice scale (4 or 6 points) only when you deliberately want to prevent a neutral dumping-ground answer and are prepared for the added respondent friction that causes.
Is a 5-point or 7-point scale better? 5-point for most general attitude and satisfaction questions — it's faster to answer and every point labels cleanly. 7-point when you specifically expect responses to cluster (most people will "agree") and you want to distinguish mild from strong agreement within that cluster.
Can a Likert scale have an even number of points? Yes — 4-point and 6-point scales exist and are legitimate when you want to force a directional lean rather than allow a neutral answer. Just make that a deliberate choice, not a default; an even scale used without intent tends to frustrate genuinely neutral respondents.
Is Likert data ordinal or interval? Technically ordinal — the gaps between points aren't guaranteed to be equal. Most teams average the numeric codes as a practical approximation anyway, which is fine at scale, but for small samples look at the full distribution of answers, not just the mean.
Surveyee's Employee Feedback and Exit Survey templates use pre-worded Likert scale questions built around exactly these principles — labeled 5-point scales, one idea per statement — as a starting point you can adjust rather than write from scratch. See our survey question types guide for how Likert scales compare to the other question formats Surveyee supports.