What is a good NPS score? How to read your own number
What is a good NPS score? Above zero means more promoters than detractors. How to work it out, how many answers you need, and what to do with the comments.
Written by OpenRated Team You sent the question. "How likely are you to recommend us to a friend or colleague?" went out to your customers on a 0 to 10 scale, the answers came back, and the tool gave you a number: 31, say. Now the obvious question: what is a good NPS score for a company like yours, and is 31 one?
The short answer is that no single number is good on its own, and the page Bain publishes on measuring the score gives none. What a score means depends on how it was worked out, how many people answered, when they were asked, and what kind of business you run. This post walks through each, with the arithmetic, so you can read your own number instead of looking up someone else's.
How the score is worked out
The method comes from Bain & Company, and its own page on measuring the score (read 2026-10-05) is the primary source for the three groups:
- Promoters scored you 9 or 10.
- Passives scored you 7 or 8.
- Detractors scored you anywhere from 0 to 6.
The score is the percentage of promoters minus the percentage of detractors. Passives count in the total but drop out of the subtraction, which is why a business where everyone answers 7 or 8 has a score of zero. The result runs from minus 100 (everyone a detractor) to plus 100 (everyone a promoter), and it is a plain number, never a percentage.
A worked example, with invented figures. Say 120 customers answered:
| Group | Scores | Answers | Share of the 120 |
|---|---|---|---|
| Promoters | 9 or 10 | 54 | 45% |
| Passives | 7 or 8 | 42 | 35% |
| Detractors | 0 to 6 | 24 | 20% |
NPS = 45 minus 20 = 25. Not 25%, and not an average of the answers, which would hide how many people are quietly unhappy. The 42 passives disappear from the number entirely, which is the method's deliberate choice and its first limit: a shop with 60% promoters and 40% detractors scores 20, and so does a shop with 20% promoters and 0% detractors, and those are very different businesses.
What is a good NPS score for a company
Most guides carry a benchmark table. Bain's page on measuring the score gives no number for "good". What it says is that the companies it calls sustained value creators, the ones with long-term profitable growth, have an NPS two times higher than the average company. That is a relative claim: good is ahead of the average for businesses like yours.
So the answer has two halves:
- Above zero means more promoters than detractors. That is the only threshold the arithmetic itself gives you. A negative score means the unhappy customers outnumber the enthusiastic ones, and whatever your industry, that needs attention before anything else.
- Good is better than your industry's average and better than your own last round. For the first, use a benchmark from a publisher that states its year, its sample (how many businesses, how many responses) and how the question was asked. Published benchmarks come from different publishers, years and samples, so this post gives none; one that cannot tell you its sample is a marketing page. For the second, you need your own history, which starts with the round you just ran.
Why a number alone misleads, in four parts:
| What moves the score besides your service | How it moves it | What to do about it |
|---|---|---|
| Industry | People answer differently about a business they have to deal with (an insurer, a utility) and one they chose for pleasure, so a score only means something next to businesses of the same kind. | Compare within your industry only, and against a benchmark that names its sample. |
| How many answered | A small count swings the score by whole points per person (the arithmetic is below). | Set a response count before you quote the number. |
| When and how you asked | A question sent the day after a delivery measures that delivery. One sent to everyone once a year measures the relationship. | Keep the trigger the same from round to round, and name it when you report the score. |
| Who answered | The customers who reply felt strongly enough to bother; the silent majority may lean either way, and you cannot know which. | Report the response rate beside the score, and chase the quiet accounts by hand. |
By industry: SaaS, healthcare, customer service and the rest
People usually ask this about one industry or one use, so take them in turn.
A SaaS or B2B software company usually has fewer customers and larger accounts than a shop. That makes the response count the first problem: a company with 80 paying accounts cannot reach 100 answers however hard it asks, and should read the comments and watch which accounts moved rather than the number. It also makes a relationship survey (everyone, once or twice a year) more useful than a survey after every support ticket.
Customer service scores are a different measurement. A question sent after a ticket closes tells you about that interaction and the agent who handled it, and it will run higher or lower than the relationship score for the same customers. Track it separately and never mix the two in one number.
Healthcare, financial services and hospitality share one thing: the moment you ask matters more than anywhere else. A patient asked while still in the waiting room, a borrower asked the week a rate went up, a guest asked at checkout after a late flight: each answers about that moment. Pick the trigger once and keep it.
Employee engagement (eNPS) uses the same arithmetic with a different question (how likely are you to recommend this company as a place to work). The response count bites hardest here: a company of 40 people where everyone answers moves 5 points when one promoter becomes a detractor. Treat an eNPS from a small team as a conversation starter.
Training and events are the clearest transactional case: the score describes the session, and says nothing about whether those attendees will buy. Use it to compare sessions with each other.
How many responses before the score means anything
Few guides work this through, so here is the arithmetic. With a given number of responses, each person is worth 100 divided by that number in percentage points. One promoter turning detractor moves the promoter share down by that much and the detractor share up by the same, so the score moves by 200 divided by the number of responses.
Take 20 responses, invented for the example: 8 promoters (40%), 8 passives (40%) and 4 detractors (20%) give a score of 20. One promoter has a bad week and answers 6 next time: 7 promoters (35%) and 5 detractors (25%) give a score of 10. One person, ten points. At 100 responses the same change is 2 points; at 400 it is half a point.
Our rule of thumb, which is arithmetic rather than a sourced standard: do not quote the score to anyone outside the business, and do not treat a move of fewer than 5 points as a trend, until a round has at least 100 responses. Below that, the comments are the result and the number is a rough sort. The same logic applies to public reviews, where a star average from a handful of ratings claims more than it can carry; how many reviews a business needs works through that side.
Response rate is the other half. Ask 500 customers and hear from 40, and the score describes those 40, who chose to reply. Report both together ("31, from 40 of 500 asked") and whoever reads it, including you in six months, knows what the figure can bear.
What to do with the follow-up comment
The 0 to 10 answer sorts people. The open comment under it ("what is the main reason for your score?") is where the content is, and it is the part most businesses leave in the tool. The score tells you how many customers are unhappy; the comments tell you why, which is the only thing you can act on.
Read the detractor comments first, every one of them, within the week the round closes. A detractor who wrote a comment has told you what to fix and is still a customer; reply personally, ask what would have made it a 9, and say what you will change. Then tag every comment with a theme (delivery time, a bug, the invoice, a named person) and count the themes. Three complaints about the same thing from 24 detractors is a fault; one each about 24 things is normal running. Fix the top theme, and ask again next round to see whether it moved.
An invented example of what the sort looks like. Ten detractor comments come in: four are about delivery time, three about invoice errors, two about one app bug, and one about a named person. The score said ten unhappy customers; the sort says fix delivery first, and the invoice next.
Two limits. Asking the question more often will not rescue a product people dislike; it will only tell you faster. And a round without the follow-up question is a number with no handle: if your survey tool cannot add an open field after the score, change the tool before you change anything else. On the customer success software listing, in-app NPS and CSAT collection is one of the rows every product is compared on; on the market research software listing, a survey builder with branching logic is, and branching is what lets a 0 to 6 answer open a different follow-up from a 9 or 10.
NPS is private feedback; a public review is a different ask
The NPS answer is yours. It goes into your tool, your team reads it, and nobody outside the business sees it unless you publish it. A public review is the opposite: a customer writes it on a platform, under their own name, where the next buyer reads it before they have ever spoken to you. Why reviews matter to a business covers what that visibility does for sales and search; the difference that matters here is who you ask.
| NPS answer | Public review | |
|---|---|---|
| Who reads it | You and your team | Anyone searching for your business |
| Who you ask | Every customer, or a defined sample | Every customer, with no filter by score |
| What you get | A trend number and private comments | Trust with people who have not bought yet |
| What it is for | Finding and fixing problems | Being chosen |
| Who controls it | You hold the data | The reviewer and the platform |
Some survey tools offer to send the 9s and 10s to a review link and the 0 to 6s to a private form. Do not switch that on. Google's help page for business owners, Tips to get more reviews (read 2026-10-05), says a mix of positive and negative feedback often feels more trustworthy, and that offering incentives in exchange for posting, changing or removing reviews is prohibited. In the United States, the Federal Trade Commission's final rule on fake reviews (announced 2024-08-14, effective 60 days after publication in the Federal Register, read 2026-10-05) bars a business from presenting the reviews on its own website as all or most of those submitted when it has suppressed reviews by their rating or negative sentiment, and bars groundless legal threats or intimidation to prevent or remove a negative review. Those are the rules as published and not legal advice; whether you can pay for reviews covers the rest of that ground.
The practical rule is simpler than the law: ask everyone who answered the NPS question for a public review, in the same message, in the same words. The promoters write the warm ones; the few detractors who write at all give the next buyer the balance that makes the warm ones believable.
The one thing to do this week
Send the next round to every customer, with the open follow-up question under the score, and read every detractor comment before the week is out. When you are ready for the public side, the same customers are the ones to ask, all of them: how to get more reviews gives the message and the timing.
