NPS Calculator

What this one actually measures

Quick clarification first, because those three letters are overloaded and you may have arrived here looking for something else entirely.

This calculates Net Promoter Score, a customer survey metric. It has nothing to do with the National Pension System, or with any pension, annuity or retirement product.

What it does is take the answers to one question. You ask customers how likely they are to recommend you to a friend or colleague, on a scale from 0 to 10. You count how many gave each score. This turns those counts into a single number between minus 100 and plus 100.

The arithmetic takes two seconds. Everything worth saying about it concerns what the number does and does not tell you, which is where the rest of this goes.

Eleven boxes, one per score

One box for each possible answer, from 10 down to 0. Enter how many people gave that score. Leave the rest blank, since empty is treated as zero.

So if 38 people said 10, 22 said 9, 19 said 8 and so on, those numbers go in the matching boxes. You are entering counts of responses, not the scores themselves.

Press Calculate and you get four counts and the score itself:

• Promoters, the nines and tens.
• Passives, the sevens and eights.
• Detractors, everything from six down to zero.
• Total responses.
• Net Promoter Score.

Having the three counts printed alongside the score is more useful than it looks, and the section on two companies with the same score explains why.

Why sevens and eights count for nothing

The formula is a subtraction expressed as a percentage:

NPS = (promoters - detractors) ÷ total responses × 100

Notice what happens to the passives. They appear in the denominator, because they are part of the total, and nowhere in the numerator. Somebody who gave you an eight has their vote counted and then discarded.

That is deliberate rather than an oversight, and it produces the metric's most argued-over feature. A seven out of ten sounds like reasonable feedback. In this scheme it is a detractor, sitting in the same bucket as a zero.

The reasoning behind the split is that the scale is not being read as a satisfaction rating. It is being read as a prediction of behaviour: only the top two boxes are thought to indicate someone likely to actively recommend you, while anything at six or below indicates someone who might actively warn people off. The middle is treated as inert.

Whether those cut points are right is a genuine and unsettled question. They were chosen from the original research rather than derived from first principles, and studies since have found that other ways of slicing the same eleven-point scale can predict outcomes at least as well.

What matters practically is that the buckets are lopsided. Seven of the eleven possible answers are detractors and only two are promoters. That is why scores that look alarming by the standards of a normal average are ordinary here, and why the range runs from minus 100 to plus 100 rather than 0 to 100.

A hundred responses, worked through

A survey comes back with 100 answers, distributed like this:

Promoters: 38 plus 22 = 60. Passives: 19 plus 11 = 30. Detractors: 5 plus 3 plus 1 plus 1 = 10.

NPS = (60 minus 10) ÷ 100 × 100 = 50.

Worth pausing on how much information that single number swallowed. Thirty people, nearly a third of your sample, contributed nothing to it except by making the denominator bigger. If all thirty had scored 8 instead of splitting between 7 and 8, the score would be identical.

And the leverage is uneven. Moving one person from an 8 to a 9 raises the score by one point. Moving one person from a 6 to a 9 raises it by two, because they leave one bucket and join the other. Your effort is worth double when it is aimed at the unhappy end.

Two companies, the same score, nothing alike

Here is the sharpest illustration of what gets lost.

Company A: 50 promoters, 0 passives, 50 detractors. NPS 0.

Company B: 30 promoters, 40 passives, 30 detractors. NPS 0.

Identical scores. Wildly different businesses.

Company A is polarising. Half its customers love it and half actively dislike it, which usually means a strong product with a serious flaw, or a service that works brilliantly for one segment and badly for another. There is something specific to find and fix, and the fix could be transformative.

Company B is unremarkable. Most people are indifferent, and the small groups at either end roughly cancel. That is a much harder problem, because there is no obvious thing to fix, only a general absence of enthusiasm.

Any sensible response to those two situations is different, and the score cannot tell them apart. Which is why the three counts printed above your score matter more than the score. Look at the shape of the distribution before you look at the summary of it.

Does it actually predict anything

This deserves a straight answer, because the metric arrived with an unusually large claim attached.

Net Promoter Score was introduced by Fred Reichheld of Bain in a 2003 Harvard Business Review article titled "The One Number You Need to Grow". The argument was that this single question predicted company growth better than any conventional measure of satisfaction or retention.

That specific claim has not held up.

In 2007 Keiningham, Cooil, Andreassen and Aksoy published a longitudinal study in the Journal of Marketing which found that the Net Promoter metric performed no better than other established measures of customer satisfaction and loyalty at predicting company growth. Later work has largely echoed that. A study comparing NPS against alternative ways of calculating a score from the same eleven-point question found that simple top-box metrics performed better, which undercuts the case for this particular arithmetic being special.

The original research has also been criticised on method, principally for using historical revenue to argue about future growth, and for resting on data gathered by the people proposing the metric.

None of which makes the number worthless, and the more recent work is more interesting than a flat verdict.

A 2021 study in the Journal of the Academy of Marketing Science found that NPS does carry predictive value for sales growth in the near term, with two important conditions. Managers get more out of tracking changes in their score than out of hitting a high absolute level. And it works when measured across all potential customers as a measure of brand health, rather than only among existing customers as a loyalty measure.

So the honest summary. It is not the one number you need. It is a reasonable, cheap, widely understood indicator that is most useful as a trend, and it should sit alongside other measures rather than replacing them. Its real advantage was never statistical. It is that one question gets answered by far more people than a long survey, and a number everybody in the building understands gets acted on.

How to read your number without fooling yourself

Watch the direction, not the level. This is the strongest finding in the recent literature and it is also the most practical. Your score against your own score last quarter is informative. Your score against a published industry benchmark is much less so, because response rates, question wording, timing and which customers were asked all move the number.

Check your sample size before you trust a decimal. The calculator reports to two decimal places, which is more precision than most surveys can support. On twenty responses, a single person changing their mind can move the score by ten points. On a small sample, round to the nearest ten in your head and treat anything smaller as noise.

Ask who did not answer. People with strong feelings answer surveys. The indifferent majority often does not, which pushes both tails of your distribution up and can make a score look more dramatic than the underlying reality.

Do not manage the number directly. Once a score becomes a target, the incentive to coach customers toward a nine or ten, to survey only the happy ones, or to time the ask carefully becomes considerable. Every one of those improves the score and none improves the business.

Read the comments. The score tells you there is a problem. The free text tells you what it is. If your survey collects only the number, you have the alarm without the address.

Questions people ask

Is this the National Pension System calculator?

No. This is Net Promoter Score, a customer survey metric. The two share an abbreviation and nothing else.

How is Net Promoter Score calculated?

Subtract the number of detractors from the number of promoters, divide by the total number of responses, and multiply by 100. Promoters scored 9 or 10, detractors scored 0 to 6.

Why do sevens and eights not count?

Passives are included in the total, which is the denominator, but not in the subtraction. The design treats them as neither likely to recommend you nor likely to warn people off, so they dilute the score without contributing to it.

What is the range?

Minus 100 to plus 100. Minus 100 means every response was a detractor, plus 100 means every response was a promoter.

What counts as a good score?

It varies enormously by industry, country and how the survey was run, which makes cross-company comparison unreliable. Your own trend over time is a far more useful benchmark than anybody else's number.

How many responses do I need?

More than most people use. On twenty responses one person can shift the score by ten points, so treat small-sample results as directional at best and ignore the decimal places.

Does a high score mean the business will grow?

Not reliably. The original claim that this was the best single predictor of growth was not supported when tested independently, and it performs about as well as other satisfaction and loyalty measures. Recent research suggests improvements in the score carry more signal than the level of it.

What is the fastest way to raise it?

Work on the detractors. Moving one person from a 6 to a 9 shifts the score twice as far as moving someone from an 8 to a 9, because they leave one bucket and join the other.

References

The metric, the promoter and detractor cut points, and the original claim that it predicts growth come from Reichheld's 2003 article. The finding that it performs no better than other satisfaction and loyalty measures at predicting growth comes from the longitudinal study by Keiningham and colleagues. The finding that changes in the score carry more predictive value than the absolute level, and that it functions better as a brand health measure across potential customers than as a loyalty measure among existing ones, comes from Lee and colleagues. The comparison against alternative calculation methods on the same eleven-point scale, which found top-box metrics performed better, comes from the study in the Journal of Business Research.

1. Reichheld, F. F. (2003). The One Number You Need to Grow. Harvard Business Review, 81(12), 46 to 54.
2. Keiningham, T. L., Cooil, B., Andreassen, T. W., and Aksoy, L. (2007). A Longitudinal Examination of Net Promoter and Firm Revenue Growth. Journal of Marketing, 71(3), 39 to 51.
3. Lee, S. et al. (2021). The Use of Net Promoter Score (NPS) to Predict Sales Growth. Journal of the Academy of Marketing Science. https://wrap.warwick.ac.uk/id/eprint/154989/1/WRAP-use-Net-Promoter-Score-(NPS)-predict-sales-growth-Lee-2021.pdf
4. Baehre, S., O'Dwyer, M., O'Malley, L., and Lee, N. Customer Mindset Metrics: A Systematic Evaluation of the Net Promoter Score (NPS) vs. Alternative Calculation Methods. Journal of Business Research. https://www.sciencedirect.com/science/article/abs/pii/S0148296322003897
5. Grigg, N. What is Wrong with Net Promoter Score. https://arxiv.org/pdf/1806.10452