How to write the NPS question correctly?

Last updated on August 21st, 2023.

11 MIN READ

This post is your ultimate guide to wording the “ultimate question”—the NPS question.

I go over the canonical way of how to ask the NPS question, and explain how to write an NPS survey question adapted for your needs.

---

Feeling impatient?

This infographic sums up the key takeaways from this article. It can serve as your template for writing NPS questions!

---

How to ask the NPS question

What is the Net Promoter Score?

To understand how to write an NPS question, it’s important to know what NPS is.

NPS stands for Net Promoter Score.
It is a proxy measure for customer loyalty.

The metric was developed and validated by the Bain consultant emeritus Frederick F. Reichheld somewhere around 2000-2002, and popularized in 2003 in a Harvard Business Review (HBR) article.

Reichheld marketed NPS as predictive of business revenue growth relative to competition in the market.

🤓 “Validated” means that the authors of the metric tested it on a lot of survey data before rolling it out for everyone else to use. In other words, many very smart people ran all sorts of statistical procedures to make sure this survey question works as intended.

And to make sure that we get the benefit of using the metric, we should follow the rules set by its creators.

In that original HBR article and in his book, Reichheld provided a set of recommendations on how to word and use the question. Throughout this article, we’ll follow his recommendations.

👉 In this article, I only cover what’s relevant for question writing and survey design when it comes to NPS. Read my other post where you can learn where the NPS metric comes from, and about some assumptions behind NPS.

What is the NPS question?

“NPS” is not the question itself. Instead, it’s the metric (“score”) we calculate using the data collected using the NPS question.

The original wording of the question that Reichheld validated was:

“How likely is it that you would recommend
[company X] to a friend or colleague?”

NPS is about likelihood to recommend

In simplest terms, NPS is a “word of mouth” score. It’s about how likely a customer of a company is to recommend it to others.

So, the necessary part of the NPS question text is:

“how likely are you to recommend … ?“.

❗There is some room for variation, but any wording of the NPS question must contain the “likelihood to recommend to others” piece.

The NPS question has an 11-point response scale

It’s best to maintain the same scale the NPS question had during development and validation.

The scale should have 11 points with the following anchors:

• 0 — Not at all likely
• 5 — Neutral
• 10 — Extremely likely

According to Reichheld’s book, we should clearly label the ends of the scale, to make the scale clearer for the respondent.

Reichheld recommends to label both “0” and “10” points with text (show “not at all likely” and “extremely likely”).

It is not too clear whether we should also label the “neutral” point, but we definitely need to understand it as such. See here for more info on whether you should label the neutral middle in the NPS scale.

Keep the other points on the scale (1, 2, 3, 4, 6, 7, 8, 9) unlabeled, that is, you should just show numbers, with no text explanation.

NPS combines responses by likelihood

The calculation process for the Net-Promoter Score is not so straightforward.

With NPS, you do not just average out all your responses.

You need to first group respondents by likelihood to recommend, into different segments.

Reichheld validated three segments that correlate with repurchase and recommendation behavior:

How scale responses correspond to NPS segments

Segment	Scores	Explanation
Detractors	0-6	Unlikely to recommend. Possible negative word-of-mouth.
Passively satisfied	7-8	Neither likely nor unlikely to recommend. They’re often called just “Passives”. I find that this paints them in an unnecessarily negative light. They are actually satisfied, even though they are not giving you a rave review. The “canonical” name is “Passively satisfied”.
Promoters	9-10	Likely to recommend. They’re very satisfied and theoretically are a free extension of your marketing department. They are spreading positive word-of-mouth, aiding business growth.

🤓 A quick note for market research analysts:

You might have the instinct to add the word “NET” to “Promoters” to differentiate your summed row on top of your cross-tab or banner. Fight this urge!

In the context of NPS, “Net-Promoter” stands for the score we calculate, so it’s more correct to label your summed rows with the names of the segment: just “Promoters”, rather than “NET Promoters”.

This way, you won’t confuse the client and will keep your consultant happy.

How to calculate the Net Promoter Score

Here’s the formula for calculating your NPS:

Net-Promoter Score =
% of promoters – % of detractors

The Net-Promoter Score is the percentage of Promoters minus the percentage of Detractors, presented as an absolute value
(so, multiply the percentage by 100 and remove the percent sign).

Very simple.

🤓 Although we use percentages to calculate NPS, we present the score as a regular number, without the percentage sign.

The resulting number can range from -100 to +100.

For example, in a sample of 1,000, you have 250 Promoters (those selecting options 9 and 10) and 500 Detractors (those selecting 0-6).

That puts your percentages at 25% promoters and 50% detractors, and your Net-Promoter Score at -25.

It makes sense that your score is negative in this case, as you have more Detractors than Promoters—so, your word of mouth is negative, rather than positive.

Surveys asking the NPS question should be short

If you look online, you can find recommendations on when to ask an NPS question inside a survey—at the beginning or at the end.

There has been research suggesting the valuations tend to be more positive when the question is upfront. There may be some interference from survey length increasing fatigue and reducing attention by the time a respondent gets to the NPS question if it’s at the end of the survey.

However, this is not the correct way to use the question.

🗯️ The metric’s author explicitly noted that we should ask the NPS question in short surveys.

Open up your e-mail inbox, and you’ll see examples of the NPS question used correctly.

Those e-mails that contain just the question—that’s the best way to go.

---

Want to learn more about NPS? 🤔

Go straight to the source. Grab the latest edition of The Ultimate Question by Fred Reichheld today.

At zero cost to you, I will earn an affiliate commission if you click through this Amazon link and finalize a purchase. Creating free resources like this one takes time. Thank you for your support! 🥰

---

How to customize the NPS question

How to target a specific interaction or transaction with NPS

Although the original (and validated) use of NPS is to evaluate the company as a whole, you may want to target a particular aspect of your company’s operations, so you can understand what impact it has on your “word of mouth” score.

The beauty—and original selling point—of NPS is its transactional nature.

NPS is less about market research insights, and more about improving business operations.

NPS was to seamlessly blend into a company’s operations and quickly help comparatively evaluate different branches, offices, field managers, or sales representatives.

After a customer completes a transaction, you can ask them the NPS question.

To target the question to a specific interaction, prepend a short qualifier to the original NPS question:

[A: Based on / Considering / Following / Thinking of]

your recent

[B: purchase / interaction with our support staff / conversation with our sales representative],

how likely is it that you would recommend [company X] to a friend or colleague?

where part A introduces the idea of remembering a past event,

and part B defines the interaction or transaction to evaluate. This can be anything relevant to your business’ operations.

How to ask an agent promoter score question

The classic NPS metric is about the customer’s overall relationship with the company.

However, it may not be as operationally specific as needed for uses.

For example, a call center will have many agents with variable performance, so it may make sense to collect a targeted score for each agent.

In this case, you may want to change the object of evaluation from the company overall to a specific product, service, or agent of the company.

For this, simply swap out “company” for something else:

How likely is it that you would recommend

[our company / brand / services / location / branch / product / this customer service representative / sales representative / technician / agent]

to a friend or a colleague?

How to adapt the NPS question to phone or CATI

To ask the NPS question over the phone (or, as we say in data collection, Computer Assisted Telephone Interviewing), append the following qualifier:

..on a scale of 0 to 10, where
0 is “not at all likely” and
10 is “extremely likely”?

It’s important to include this wording when administering the NPS question over the phone to let the respondent know what the numeric response scale is.

Otherwise the respondent will feel lost, you’ll lose expensive calling time, and you won’t produce valid data. A phone survey respondent should be able to refer to the correct scale just as a web respondent would by looking at the screen.

How to change the response scale for the NPS question

Should you label the “neutral” middle?

Recommendations from Reichheld are not clear on whether the middle option “5” should be labeled, but he does consider it as the “neutral” option.

Some research suggests that labeling the neutral midpoint has no statistically significant effect when you ask the NPS question to brand customers, but it does have an effect when non-customers are asked the question.

In any case, since NPS is a measure of customer loyalty… you should technically be asking the NPS question only to the customers of a brand.

So, it’s mostly up to your personal preference, or, if you are tracking the results, maintaining how the question was asked historically.

⚠️ PROCEED WITH CAUTION.

Originally, Reichheld validated NPS with 11 response selections (from 0 “not at all likely” to 10 “extremely likely”, with a “neutral” middle at 5).

Before shortening or in any way adjusting the response scale, make sure to consult with your firm’s resident statistician or data scientist. Plan how you’ll calculate your Net-Promoter Score given the change in scale.

How to shorten the NPS scale

If you absolutely must shorten the scale—perhaps you have screen limitations where you need to display the whole scale horizontally on a mobile phone—choose a scale with an odd number of selections.

You’ll need an odd number of options to maintain equal gaps between the extreme anchors and the middle. Like in the original 11-point scale, which has 4 points between the extremes and the middle, you’ll want to maintain a consistent number of points in an alternative scale.

Realistically, you only have two options other than the 11-point scale:

• 7-point scale with a 2-point gap:
  1 — Not at all likely | 2 | 3 | 4 — Neutral | 5 | 6 | 7 — Extremely likely
• 5-point scale with a 1-point gap:
  1 — Not at all likely | 2 | 3 — Neutral | 4 | 5 — Very likely

Ideally, you’ll want to still have accurate text labels for the extremes and the middle.

Why not shorten the NPS scale?

Here’s how shortening the scale could invalidate your scores:

• with both a 7- and 5-point scale, you lose that “0” point, which is a representation of a really bad customer experience;
• with a 5-point scale, you lose the extremes at the ends of the scale;
• your calculation of segments (Promoters, Passively satisfied, and Detractors) will be different and not as originally intended.

Importantly, Reichheld strongly recommended to keep the NPS question to an 11-point scale including a zero. For example, he has some arguments on how this scale is best for international applications. Be sure to read the second edition of the book on NPS to learn more.

How to not misrepresent the scale options

Notice that the question template in the infographic highlights only 0, 5, and 10 points of the scale. The respondent should only feel visually anchored to the extremes and maybe the middle of the scale.

❗Beware of this when referencing online sources. Many companies, particularly in the research tech space, create promotional blog posts about NPS. Those posts often contain infographics explaining how to present the NPS question. I’ve seen some posts that highlight points corresponding to NPS segments (0-6, 7-8, 9-10) right inside the survey question. Those segment point groupings are for the researcher, not the respondent! If you don’t know NPS well, you might actually end up programming your question in a way that highlights this division of numbers. Your respondent shouldn’t feel compelled to click on the “green part” of the scale highlighting options 9 and 10.

How to ask NPS follow-up questions

Although NPS is a validated metric, the numbers will not provide the reasons for the rating or explain what the organization needs to improve.

For this reason, we should include open-ended follow-up questions.

You can do it in two ways: a default for all respondents, and tailored to NPS segments.

How to ask a simple open-ended NPS follow-up

A common way to phrase an NPS follow-up question is:

What is the (primary) reason for your rating?

Short and sweet.

If you’d like to prompt your respondent to only leave one reason, include the word “primary”. This will make it easier both on your respondent, and on you for coding the responses. (It’s less effort to code open-ended responses on one variable rather than multiple variables.)

How to tailor an NPS follow-up to different segments

Why tailor follow-up questions?

In the original article on NPS, Reichheld recommended to tailor your follow-up questions to each audience (Promoters, Passively satisfied, and Detractors).

Primarily based on their customer experience, each of the audiences is likely to have a different attitude if questioned further:

• Detractors are more likely to have had a bad experience and be upset.
• The Passively satisfied might not have much more to say.
• Promoters might provide some positive feedback that specifically highlights what you’re doing well.

Since some respondents will like you more than others, questions about how they feel about you might elicit different feelings. So, it’s a good idea to speak to them differently.

Why Promoters stand against Passives and Detractors

Your operational goal will be, typically, to increase your NPS score: that is, making sure more people rate at 9 or 10 rather than 0-8.

In other words, your goal as a business is to make both Detractors and the Passives into Promoters. Due to this, you may find yourself using the same line of questioning for both the Passively satisfied and Detractors. It might be more efficient to have the one follow-up question for Promoters, and the same follow-up questions for the Passively satisfied and Detractors.

Examples of NPS follow-up questions by segment

Here are some examples of follow-up questions you might want to ask Promoters, the Passively satisfied, and Detractors:

NPS segment	NPS follow-up question text
Promoters	Would you like to tell us what exactly excites you about us? What do you like about us the most? Which features of the product do you value the most? How did we meet your needs today?
the Passively satisfied	What is the one thing we could improve to make your experience better? What could we (is the one thing we could) do to make you happier? Which features of the product do you like the least? What can we do to get a rating closer to 10? What is the primary reason why you would not recommend us? What is missing in your experience with us?
Detractors	We’re sorry to hear that. What could we (is the one thing we could) improve to make your experience better? We always want to make things right. What could we (is the one thing) we could do to make you happier? We’re sorry to have let you down. How can we make it right? [You could ask this version only to the very low scores, e.g. 0-3] What can we do to get a rating closer to 10? What is the primary reason why you would not recommend us? What is missing in your experience with us?

How to limit the amount of text respondents enter in NPS follow-up questions

From my own experience working on surveys, the most unsatisfied respondents will have a lot to say.

For you as a researcher that’ll mean a lot of text to sort through.

To try and limit the entries, you may want to include prompts like “the one thing we could do” or “primary reason why” when wording your questions to prompt your respondents to leave just one reason—of course, if that is your preference.

This will make it easier to code open-ended responses (verbatims).

---

Hope this helps. Thanks for reading!

📝 Eager to put this into practice?

Go ahead and learn how to create an NPS survey in Google Forms for free!