Employee Net Promoter Score (eNPS)

What the employee Net Promoter Score (eNPS) is

The employee Net Promoter Score, or eNPS, is a figure drawn from a single question put to employees: how willing are you to recommend working at this organisation, on a scale from 0 to 10?

The answers are then sorted into three categories. Anyone who answers 9 or 10 is a promoter, anyone who answers 7 or 8 is a passive, and anyone who answers 6 or below is a detractor. The score is the percentage of promoters minus the percentage of detractors, and it ranges from minus 100 to plus 100. That split into promoters, passives and detractors is the one used in every example below; we do not say that it is the correct split, or the only one.

The first thing to know about the score is that it is not a percentage, even when it is written like one. It is the difference between two percentages, and there is nothing of which a score of 25 is 25 percent. Read it as a percentage and it appears to say that a quarter of employees are satisfied, which it does not say.

A full eNPS calculation

Consider 200 respondents divided as follows: 90 promoters, 70 passives and 40 detractors.

• Promoters: 90 out of 200, which is 45 percent.
• Passives: 35 percent.
• Detractors: 20 percent.
• The score: 45 minus 20, which is plus 25.

Notice that 35 percent of the respondents do not enter the subtraction at all. The passives count in the denominator but do not appear in the numerator, and that is the first reason the score cannot be taken at face value.

Whether plus 25 is a good result is a separate question. We have no published source for the Saudi market on which to base a figure for what counts as a good eNPS, or a threshold for an acceptable one, and we draw no conclusion from that absence.

How the eNPS categories discard the detail in the answers

Turning eleven points on the scale into three categories throws away detail the answers carry, and a simple comparison shows the effect.

• A first organisation: all 200 of its employees answer 8. The average is 8.0, there are no promoters and no detractors, and the score is zero.
• A second organisation: 120 answer 9 and 80 answer 6. The average is 7.8, promoters are 60 percent and detractors 40 percent, and the score is plus 20.

The second organisation has the lower average and a score 20 points higher. There is no arithmetical error in that. The score does not measure the average; it measures how far answers sit at the two ends. An organisation where nobody is discontented and nobody is enthusiastic scores zero, while one whose people have split into two camps scores above zero.

So the score is examined alongside the distribution of answers, not on its own. A single figure can stand for two situations that are not managed in the same way: general indifference, and sharp division.

Three distributions with the same eNPS

Because the score is a difference between two percentages, one difference can come from more than one distribution. Here are three organisations with 200 respondents each, and every one of them scores plus 25:

• The first: 90 promoters, 70 passives and 40 detractors, or 45, 35 and 20 percent.
• The second: 60 promoters, 130 passives and 10 detractors, or 30, 65 and 5 percent.
• The third: 120 promoters, 10 passives and 70 detractors, or 60, 5 and 35 percent.

These three are not managed in the same way. In the second, 65 percent of people sit outside both ends, which describes an organisation whose people neither complain nor show enthusiasm. In the third there are 70 detractors, more than a third of respondents at the negative end, and that is a situation in which a decision is needed now.

Yet the score is identical, and a dashboard that shows it alone presents all three as the same thing. That is why the least that should be shown with it is the number of detractors: that is the figure on which action is based, and the composite score does not carry it.

How many people it takes to move eNPS by ten points

In a sample of 200, each respondent is worth half a point in either percentage. It follows that one person’s change of answer has a different effect depending on the category the person leaves and the category the person joins:

• A detractor who becomes a promoter: the promoter percentage rises by half a point and the detractor percentage falls by half a point, so the score moves by a full point.
• A detractor who becomes a passive: the score moves by half a point.
• A passive who becomes a promoter: the score again moves by half a point.

Reaching ten points therefore takes 10 people moving from the lowest category to the highest, or 20 people moving one category each. The first route needs fewer people but is harder in practice, because it requires someone who was discontented to reverse their view.

What makes this calculation useful is that it turns a target such as “raise the score by ten points” into a statement that can be tested: 20 employees out of 200 need to change their position during the period. If nothing in the plan can be expected to do that for twenty identifiable people, the target is a number written down with no intention behind it.

What the cutoff at 6 does to eNPS in a small team

For a single respondent, the difference between a 6 and a 7 is one step on a scale of eleven points. Its effect on the score is a move from the detractor category to the passive category, which changes completely how that answer is counted.

Among 200 respondents, that step is worth half a point of the score, an effect the figure can absorb. In a team of 10 it is worth 10 full points, because each respondent is a tenth of the sample.

The score of a team of ten can therefore only take values that sit 10 points apart, so a claim that it improved from 20 to 30 has no meaning: that is the smallest change possible, and a single person moving an answer by one step can produce it. For that reason the score is not shown at the level of small teams, and no decision is based on its movement there.

What the employees who did not respond mean for eNPS

The score is calculated on those who answered, while the decision affects everyone. If an organisation has 500 employees and only 200 of them responded, the score of plus 25 describes the respondents.

Were the remaining 300 all assumed to be detractors, promoters would be 90 of 500, or 18 percent, detractors 340 of 500, or 68 percent, and the score minus 50. At that extreme the score falls 75 points below the one reported.

The extreme is not a likely case, and setting it out is not a forecast. The point is that the response rate is part of the score, not a piece of information beside it, and a score built on 40 percent of employees should say so on the same line. This matters more because those who do not respond may not be a random sample.

Why eNPS is not averaged across teams

Suppose a team of 10 has 8 promoters and 2 detractors, which gives a score of plus 60, and a team of 100 has 5 promoters, 5 detractors and 90 passives, which gives a score of zero.

The average of the two scores is 30, while the score calculated on all 110 people together is 5.45. The gap between the two figures is more than 24 points.

So the score is neither added up nor averaged; it is recalculated on all the answers. A dashboard that shows the average of team scores shows a figure that belongs to nobody, and its error grows as the teams differ more in size.

What eNPS gets mistaken for

• Employee engagement. Engagement describes a state that is measured with several instruments and examined at the level of the team, whereas eNPS is one number from one question. The score may be an effect of engagement without being a measure of it.
• A pulse survey. This is a short, repeated instrument, and the eNPS question may or may not be among its questions. The survey is the container; eNPS is a measure that can be placed in it or in something else.
• Employee referral. A referral is something that has actually happened and is counted by number, whereas eNPS records a stated willingness, measured by a question. An employee who answers 10 and refers nobody counts towards the score and does not appear in the referral count.

We have no published measurement of the relationship between eNPS and employees leaving, whether leaving is tracked as attrition or through a retention rate, and none of the figures above addresses it.

What distorts a reading of eNPS

• Showing it without a response rate or a count of respondents. As shown above, a missing response rate leaves a wide range of possible values open, and the count of respondents sets the smallest movement the score can make.
• Measuring it straight after an event. A score collected in the week bonuses are announced differs from one collected in the week an organisational change is announced, and the difference between them says nothing about the organisation.
• Tying it to a manager’s bonus. The question then turns from a measurement into a request, and the score becomes higher and less honest at the same time.
• Comparing it with other organisations’ scores. The categories themselves are applied in different ways, the question is worded differently, and culture affects how people use the two ends of the scale.
• Relying on it without an open question. The score says where people stand but not why, and a single line of explanation is more useful than following the score for four quarters. The open question records why respondents answered as they did, much as an exit interview is held to learn why an employee is leaving.

The wording of the question and the method of collecting answers are fixed before the first measurement, because changing a single word, or changing the channel through which answers are gathered, breaks the time series without anything showing in the figures. A series whose basis has broken gets interpreted as a change in the people when it is a change in the instrument. How often to measure follows from what the organisation does with the result, and we set no interval for it.

Before an eNPS figure is reported

Four things are shown with the score: the number of respondents, the response rate, the distribution across the three categories rather than the difference alone, and the date of collection. The distribution is what separates indifference from division, and the score by itself cannot tell them apart.

Decisions rest not on the score but on what people say in the open question that follows it. The score can signal that something has changed; it cannot say what. An organisation that tracks it for four years with no second question will have collected a precise time series about something it cannot identify.

This is an explanation of the concept, not legal advice.

Related terms

• Retention Rate
• Exit Interview
• Employee Value Proposition
• Attrition
• Graphic Rating Scale