What is NPS? The Net Promoter Score explained

Where NPS comes from

The Net Promoter Score was introduced in 2003 by Fred Reichheld of the consultancy Bain & Company, in a Harvard Business Review article with the telling title The One Number You Need to Grow. The idea: instead of long satisfaction questionnaires, ask a single question that says something about loyalty. Recommending a company puts your own reputation on the line, and people only do it when they're genuinely happy.

Many organisations now use NPS as a standing customer metric. Net Promoter, NPS and Net Promoter Score are, incidentally, registered trademarks of Bain & Company, NICE Systems and Fred Reichheld.

The NPS question

Bain calls it "the ultimate question":

How likely are you to recommend [company] to a friend or colleague?

Customers answer with a number from 0 (not at all likely) to 10 (extremely likely), which places them in one of three groups:

Bain describes the groups roughly like this. Promoters are loyal, enthusiastic customers who recommend you and buy more over time. Passives are satisfied for now, but recommend you less and are more likely to defect to a competitor. Detractors are unhappy, spread negative word of mouth and churn more often.

How to calculate NPS

A worked example

An online shop asks customers for their NPS a week after delivery, and 250 customers respond:

The NPS is 44 − 22 = +22. Passives count towards the total of 250, but not in the subtraction. The NPS calculator checks your sums in seconds, even from a pasted column of individual scores.

How precise is that number?

An NPS based on 250 responses is an estimate of what all your customers think. At 95% confidence, the margin of error in this example is about ±10 points: the true NPS is likely somewhere between +12 and +32. A dip to +18 next month is no reason to panic. The calculator always shows that margin alongside the score.

Relationship NPS and transactional NPS

In practice NPS is measured in two ways, and it helps to keep them apart:

• Relationship NPS. You ask customers periodically, say quarterly or annually, how they feel about you overall. It reflects the relationship as a whole and is the measure to track as a trend.
• Transactional NPS. You ask shortly after a specific experience: a delivery, a call with customer service, a first month in your app. It says more about that moment than about the relationship.

Both are useful, but don't blend them into one number. A customer who has just reported a problem will score differently from the same customer three months later. To measure one specific experience, CSAT or the Customer Effort Score is often the better fit. We compare them in NPS vs CSAT vs CES.

What is a good NPS?

Anything above zero means you have more promoters than detractors. Beyond that, "good" is mostly relative. NPS varies widely between sectors: an energy supplier and a bookshop start from very different places. Timing and method matter too. An NPS asked straight after a successful purchase can come out very differently from one sent once a year to your entire customer base.

That leaves three comparisons worth making:

1. With yourself, over time. Same question, same kind of customer, same moment. This is the one that matters most.
2. Between your own segments. Customers of branch A against branch B, new customers against long-standing ones. Big gaps point the way.
3. With competitors in your sector, but only if you're sure they measure the same way. Published benchmark tables vary widely, and it's rarely possible to see how they were put together.

Pitfalls

Only reaching happy customers. If you only ask customers who completed an order, you miss everyone who gave up halfway. Think about who never sees your question.

Staff asking for a 10. Hold a team to account for its NPS and you'll soon hear "you'll get a short survey, a 10 would be lovely". That makes the number worthless.

Calculating averages. A mean score of 7.6 isn't an NPS and tells you something different. The strength of NPS lies in the distribution: how many people are genuinely enthusiastic, how many genuinely unhappy.

Forgetting the passives. They don't count in the subtraction, but they're the customers most easily won over by a competitor. A large group of passives is an opportunity, and a risk.

The number without the why. An NPS of +22 doesn't tell you what was bothering the 55 detractors. Without an explanation you end up guessing at causes, and may fix the wrong thing.

From score to action

NPS becomes valuable once you know why people gave the score they did. The usual approach is an open follow-up question: what is the main reason for your score? Then comes the work: reading the answers, grouping them into topics and seeing which topics come up differently among detractors and promoters.

A few habits that help:

• Read detractors and promoters separately. What detractors raise tells you what's going wrong. What promoters raise tells you what to protect.
• Close the loop. In customer research where people aren't anonymous, you can contact a detractor personally, provided you said beforehand that you might. It solves their problem and often yields the most concrete story.
• Count topics, not words. "Delivery", "shipping" and "parcel came late" are the same thing. Group first, then count.

Example questions for a complete customer satisfaction survey are in our guide to customer satisfaction survey questions.