How to calculate NPS (and what a good score is)

The arithmetic of Net Promoter Score takes one line. The hard part is everything around it: which question you ask, when you ask it, who answers, and what you do with the people who gave you a 3. A clean formula on a badly run survey gives you a precise number that means little. This guide covers the formula briefly, then the parts that decide whether your score is worth reporting.

The formula, in short

Ask "How likely are you to recommend us to a friend or colleague?" on a scale from 0 to 10. Answers of 9 or 10 are promoters, 7 or 8 are passives, and 0 to 6 are detractors. NPS is the percentage of promoters minus the percentage of detractors, so it runs from -100 to +100.

With 200 answers, 110 promoters, 50 passives and 40 detractors give 55% minus 20%, which is +35, with a margin of error of about ±11 points. Our free NPS calculator does this for your own counts or raw scores, in your browser.

What a good NPS is

Any score above 0 means you have more promoters than detractors. Past that, published averages are hard to use, because the score moves with things that have nothing to do with how good you are:

• The moment you ask. A survey right after a solved ticket and a yearly survey about the whole relationship measure different things.
• Who answers. Customers who just had a problem answer more often than customers who are quietly happy, or the other way around, depending on how you send the survey.
• The channel and the wording. An email, an in-app prompt and a question at the end of a chat all collect different scores, even from the same customers.
• Sample size. With 200 answers the margin of error is about ±11 points, so two companies 11 points apart may not differ at all.

That is why the most reliable comparison is with yourself. Keep the question, the moment, the channel and the audience the same, and a change in the score reflects a change in your customers' experience, not in your method. A good NPS, in practice, is one that is higher than your own last score by more than the margin of error.

Design the survey

Keep the standard question

Use the standard question and the full 0 to 10 scale. Label the ends, for example "0, Not at all likely" and "10, Extremely likely", and nothing in between. If you change the wording or shorten the scale, your score stops being comparable with your own past scores.

Add one open question

The number tells you how many detractors you have. The comment tells you why. Ask one follow-up: "What is the main reason for your score?". One open question gets more answers than three, and "main reason" pushes people to name one thing you can act on.

Don't lead

Never ask for a 9 or 10, and never tell staff that their pay depends on the score. Both push the number up and hide the problems you need to see. Keep the survey short: the NPS question, the open question, and nothing that a customer has to think hard about.

Choose when to send it

• Pick one moment and keep it. If you sometimes send after a ticket and sometimes once a quarter, the trend mixes two different scores.
• Send while the experience is fresh. A survey about a ticket that arrives a week later measures memory, not service.
• Limit how often one person is asked. Customers who get a survey after every message stop answering, and the ones who keep answering are not typical. Set a rule, such as no more than one survey for each person each quarter.
• Ask everyone in the group, not a chosen few. Sending only to customers your team picks is a quick way to a score that means nothing.

Follow up with every group

A survey that nobody answers back teaches customers that the survey is a formality. Decide the follow-up for each group before you send the first survey:

• Detractors (0 to 6): a person reads every comment and replies, with a target time your team can keep, for example within two working days. Ask one question, fix what you can, and say what changed.
• Passives (7 or 8): read the comments in bulk. Passives often name the one thing that would make them promoters.
• Promoters (9 or 10): thank them. This is also the right moment to ask for a review or a referral, because they just told you they would recommend you.

Reply to a detractor like a person, not a template. If you need a starting point, the canned responses library has replies for complaints and escalations that you can adapt.

Act on detractors first

Fewer detractors often moves the score more than new promoters. Take the example above: 110 promoters, 50 passives and 40 detractors, an NPS of +35.

• Fix the problem for 20 detractors so that they become passives, and the score rises to +45.
• To reach the same +45 by adding promoters alone, you need 37 new promoters and no new detractors.

To find which problems to fix:

1. Export the detractor comments for the period.
2. Tag each comment with one cause, such as slow reply, wrong answer, product bug, price or delivery. Keep the list of causes short.
3. Count the comments for each cause. A few causes usually explain most of the comments.
4. Fix the top cause, then measure again with the same survey.

If the top cause is "could not find an answer", look at the questions behind it. A repeat question with a good answer in your help center is a findability problem. A repeat question with no answer anywhere is a missing page.

Report the score honestly

• Report the number of answers and the margin of error with every score.
• Show the trend over at least a few periods, not one number.
• If the change from last period is smaller than the margin of error, say that the score did not change. Our NPS calculator shows the margin for your own counts.
• Put the top three detractor causes next to the score, with what you are doing about each.

NPS works best next to other measures. To score a single support interaction instead of the whole relationship, use the CSAT and customer effort calculator.