Every rating scale is a trade between two things that pull in opposite directions: granularity (how much distinction the scale can capture) and ease of use (how quickly and confidently a respondent can pick a point on it). Pick too few points and you flatten real differences into noise; pick too many and respondents start guessing at the difference between a 6 and a 7. Here's how to choose.
What granularity actually buys you
A 5-point scale gives respondents five buckets to sort themselves into. A 10-point scale gives them ten. More buckets means, in theory, more precision — you can distinguish "mildly satisfied" from "very satisfied" instead of lumping both into "satisfied."
In practice, that precision only pays off if respondents can reliably place themselves on the finer scale. Most people can meaningfully distinguish about 5-7 levels of a subjective feeling — beyond that, the extra points add apparent precision without adding real signal, because respondents start picking somewhat arbitrarily between adjacent numbers that don't feel meaningfully different to them. This is a long-standing finding in survey methodology, not a Surveyee-specific claim: it's why most established scales top out around 7 points, with the notable exception of NPS.
5-point scales: the default for most attitude and satisfaction questions
A 5-point Likert scale (Strongly disagree → Strongly agree, or an equivalent satisfaction range) is the right default for most attitude, agreement, and satisfaction questions. It's fast to answer, easy to label fully (all five points can have a clear word, not just the ends), and produces data that's simple to summarize and compare across questions.
Use a 5-point scale when:
- You're measuring agreement with a statement or satisfaction with an experience, and you want data respondents can answer quickly without deliberating.
- You plan to ask many similar statements in a row (an engagement survey with 15 Likert items) — consistency and speed matter more than fine-grained precision at that volume.
- You want every point clearly labeled, not just the ends, which becomes harder to do cleanly past 5-7 points.
We go deeper on wording and analyzing this format in our Likert scale guide.
7-point scales: when you expect answers to cluster
A 7-point scale is worth the extra granularity specifically when you expect most responses to bunch up in one region of the scale, and you want to distinguish "mild" from "strong" within that cluster. If you already know most customers will say they're "satisfied," a 5-point scale collapses the interesting variation (are they barely satisfied or very satisfied?) into one bucket, while a 7-point scale preserves it.
The cost is real, though: a 7-point scale is slower to label fully and asks slightly more of the respondent's attention per question. It's a reasonable trade for a smaller number of important questions, less so for a long battery of items where speed matters more.
10-point / 0-10 scales: NPS, and not much else
The 0-10 scale is best known through Net Promoter Score ("how likely are you to recommend us, 0-10"), and it's worth understanding why NPS specifically uses 11 points when most other scales don't:
- NPS deliberately uses the extra granularity to define three named groups — detractors (0-6), passives (7-8), promoters (9-10) — and the score is calculated from those bucket boundaries, not from the raw mean. The scale's width exists to support that categorization, not because respondents can meaningfully distinguish an 8 from a 9 in isolation.
- It's a single, standardized, benchmarked question. Because NPS is asked the same way across companies and industries, its specific 0-10 format is part of what makes it comparable — changing it to a 5-point scale would break the comparison to published NPS benchmarks, even if a 5-point version would be easier to answer.
- It's asked rarely, not repeatedly. NPS is typically a single question asked quarterly or after major milestones, not one of fifteen items on a page — so the small extra cognitive load of a wider scale matters less than it would in a long Likert battery.
Outside of NPS and a handful of other standardized instruments, a 0-10 or 1-10 scale for a general satisfaction or agreement question is usually more granularity than the question needs, and it costs you labeling clarity (few teams label all 11 points, so respondents are picking a number against only two anchors). If you're building a general satisfaction question rather than implementing NPS specifically, a 5- or 7-point scale will usually serve you better. See our comparison of NPS vs. CSAT vs. CES for when the 0-10 NPS format is actually the right tool versus when a simpler CSAT scale fits better.
What each of these questionnaire scale types actually looks like
It's easier to choose a scale by seeing the actual respondent-facing wording, not just the abstract point count. Here's how the same underlying question — asking how strongly someone agrees, or how likely they are to recommend — reads at each width:
5-point (labeled agree-disagree scale): "I would recommend this product to a colleague." Strongly disagree · Disagree · Neutral · Agree · Strongly agree
7-point (labeled agree-disagree scale): "I would recommend this product to a colleague." Strongly disagree · Disagree · Somewhat disagree · Neutral · Somewhat agree · Agree · Strongly agree
0-10 (the standard NPS question format): "On a scale of 0-10, how likely are you to recommend [company] to a friend or colleague?" 0 – Not at all likely … 10 – Extremely likely (only the two ends are labeled; 1–9 are left as bare numbers — that's the standard format, not a shortcut someone skipped)
Notice the NPS question isn't just "the 5-point question stretched to eleven points" — it's phrased differently (a likelihood question, not an agreement statement) and only labels the endpoints. That's exactly why NPS stays easy to answer despite having more points than a 7-point Likert scale: respondents aren't interpreting nine unlabeled degrees of agreement, they're placing a likelihood estimate between two clear anchors.
The even-vs-odd debate: should there be a neutral midpoint?
An odd-numbered scale (5, 7) has a true middle point — "neutral," "neither agree nor disagree." An even-numbered scale (4, 6) has no middle, forcing respondents to lean one direction or the other.
The case for odd scales (with a midpoint): some respondents genuinely are neutral, undecided, or the statement genuinely doesn't apply to them. Removing the midpoint doesn't make that respondent suddenly have an opinion — it just forces them to pick a side that doesn't reflect how they actually feel, which adds noise rather than removing it. For most general-purpose attitude and satisfaction surveys, we recommend keeping the midpoint.
The case for even scales (forced-choice, no midpoint): when you specifically want to know which direction someone leans and a "neutral" option would just become a dumping ground for anyone who doesn't want to think hard about the question, removing the midpoint can produce more decisive data. This is a legitimate, deliberate design choice — but it should be a choice, not an accident. An even scale used without that intent tends to frustrate genuinely neutral respondents, who either abandon the question or answer inconsistently rather than commit to a side they don't hold.
What this looks like in practice: take the statement "The onboarding process was easy to follow."
- 5-point (with midpoint): Strongly disagree · Disagree · Neutral · Agree · Strongly agree — a respondent who genuinely found onboarding "just okay, neither easy nor hard" has a truthful option to pick.
- 4-point (forced-choice): Strongly disagree · Disagree · Agree · Strongly agree — that same respondent now has to round to a side. Some will round toward "agree" out of politeness bias (a well-documented tendency to lean positive under uncertainty), which can quietly inflate your scores rather than genuinely resolve the ambiguity.
This is why forced-choice scales work best for questions where you specifically want to know which way someone leans even on a weak preference (e.g., "would you rather X or Y"), rather than for general satisfaction measurement, where a genuine "neither" is common and worth capturing accurately.
Our advice: default to an odd scale with a labeled midpoint unless you have a specific reason to force a lean, and if you do remove the midpoint, be aware you're trading a small amount of respondent friction for more decisive (but not necessarily more accurate) data.
FAQ
What is a good NPS question format? The standard, benchmarked wording is: "On a scale of 0-10, how likely are you to recommend [company/product] to a friend or colleague?" with only the two endpoints labeled (0 = Not at all likely, 10 = Extremely likely). Don't relabel every point or shorten the scale — deviating from the standard format breaks comparability with published NPS benchmarks.
Is a 10-point scale too granular for a general survey? Usually, yes, outside of NPS specifically. Most people can only meaningfully distinguish about 5-7 levels of a subjective feeling, so a 0-10 scale for a general satisfaction or agreement question tends to produce more apparent precision than real signal, and it's harder to label fully.
Should every question in a survey use the same scale length? Within a single survey, yes — mixing a 5-point scale for some questions and a 10-point scale for others makes results harder to compare and makes the survey feel inconsistent to respondents. Across different surveys, it's fine, and often correct, to use different scales for different purposes: 5-point Likert scales for an engagement survey, the standard 0-10 for a separate quarterly NPS survey.
Does removing the neutral midpoint actually get better data? It gets more decisive data, not necessarily more accurate data. Removing the midpoint forces genuinely neutral respondents to lean one way, which can inflate whichever direction people default to under uncertainty (often mildly positive, due to politeness bias) rather than reflecting their true position. Use a forced-choice scale deliberately, not as a default.
Choosing in practice
A quick decision guide:
| Situation | Recommended scale |
|---|---|
| General attitude/satisfaction question, want it fast to answer | 5-point, odd, labeled |
| Expect answers to cluster and want to distinguish intensity within that cluster | 7-point, odd, labeled |
| Implementing NPS specifically | 0-10, standardized — don't modify |
| Want to force a directional lean, no neutral option | 4- or 6-point, even, deliberate |
Whatever you choose, keep it consistent within a single survey — mixing a 5-point scale for some questions and a 10-point scale for others on the same page makes the survey feel unpolished and makes it harder for you to compare results across questions later. Surveyee's Customer Satisfaction template uses a consistent star (5-point) scale throughout, and the NPS Survey template keeps the standard 0-10 format intact rather than reinventing it.