What Net Promoter Score Actually Measures: How It Is Calculated, How to Ask, and How It Relates to Word of Mouth

Plenty of brands adopt NPS because they want to know whether customers are satisfied, put the score in the monthly report, and then nothing further happens.

There is nothing wrong with the metric itself. The problem is that it gets treated as a report card rather than a diagnostic tool. If the score moves up or down and nobody can say why, that table is just reassurance.

This piece is about what NPS actually measures, what it cannot measure, and how to make it genuinely connect to word of mouth.

How the score is calculated

You ask one standardized question: how likely are you to recommend us to a friend or colleague? Answer on a scale of zero to ten.

Respondents are then split into three groups:

Score	Group	What it means
Nine to ten	Promoters	Clear intent, most likely to speak up unprompted
Seven to eight	Passives	Satisfied but unmoved, and replaceable at any time
Zero to six	Detractors	An unresolved problem or a negative feeling

The score is the share of promoters among all respondents minus the share of detractors among all respondents, with passives excluded from the numerator. The result falls between minus one hundred and plus one hundred.

Two deliberate trade-offs in that design are worth noticing:

• Passives are left out of the numerator. This makes the metric more sensitive to change, at the cost of discarding information about an entire group — often the group easiest to move.
• A seven and a zero land in different groups, while a zero and a six land in the same one. The grouping is itself an act of compression, and the score alone cannot reconstruct the original distribution.

The conclusion: the number is good for reading trends and poor as a conclusion. Without a follow-up question behind it, there is very little it can answer.

A score is not recommending behavior

This is the limitation of NPS that most needs to be understood.

The survey asks about intent. Word of mouth requires behavior. Several thresholds sit between the two:

• Someone willing to recommend you may never encounter an appropriate occasion to do it
• Someone may have the intent but be unable to describe you, with nothing concrete to point at
• Someone may have the intent and the words, but no channel or material to share
• Someone may actually recommend you to a person who has no need at that moment

The reverse also holds: a low score does not mean there is no word of mouth. It may just be that respondents happened to cluster among the people who hit a problem. Survey samples are never random — the people willing to spend time answering are usually at one of the two extremes, particularly satisfied or particularly unhappy.

The more reliable reading is to treat NPS as an internal trend signal while tracking real external evidence alongside it:

• The volume, content, and movement of reviews on platforms
• The character of the discussions that surface on a brand-name search
• The share of new customers who arrived through a referral
• Whether the types of question reaching support have changed

For how to set up tracking on external indicators like these, see How to Measure Word-of-Mouth ROI.

Survey timing and channels

Timing matters more than the questions. The same survey sent at different moments returns entirely different things.

Moments worth sending

• Shortly after a customer has used the product or service fully for the first time, while the impression is still vivid
• At the close of a service cycle — a project wrapping up, a course of treatment ending, a program finishing
• After a problem has been properly resolved, a moment often overlooked; the emotional turn has just happened and willingness to respond is usually higher
• Around the decision to renew or repurchase

Moments to avoid

• Before the transaction is complete, or while the customer is waiting on an outcome
• Immediately after a negative incident, when what you collect is emotion rather than diagnosis
• When the same customer has already been surveyed recently

Choosing the channel

Channel	Fits	What to watch
Email	B2B, subscription services	Low response rates, needs a clear subject line
Messaging apps	Consumer services with an existing interaction	Avoid feeling intrusive; keep frequency restrained
In-product or on-site	Digital services	Do not block whatever the user is in the middle of
On location or on paper	In-person services	Answering face to face skews positive
Phone interviews	High-value, low-volume customers	Expensive, but the highest information density

The bias in answering on location deserves particular attention: it is hard to give a low score honestly while standing in front of the person who served you. Scores collected in that setting usually run high and are not suitable to place side by side with online survey data.

The open-ended question is the point

If the survey could keep only one question, keep the open-ended one, not the rating.

The standard practice is to follow the rating with one line: “What is the main reason?” The answers to that question are the material you can actually change things with.

The follow-up can be adjusted by group:

• For promoters: what specifically makes you willing to recommend us? These answers are the best source of word-of-mouth material and the basis for the wording of review requests; for the method, see How to Ask Customers for Reviews.
• For passives: what would we have to do before you would recommend us unprompted? This group’s answers are usually the most specific and the easiest to act on.
• For detractors: which part of the experience left you feeling that way? This is the most valuable question of all, on the condition that you genuinely intend to go back and deal with the answer.

Turning detractor feedback into a fix list

The worst thing to do with a low score is leave it in a report as a number.

A workable process

1. Classify by stage, not just severity. Sort feedback by where it happened: pre-sale information, the purchase flow, the product itself, delivery, after-sales, support manner. Once sorted, you can see where the problems concentrate.
2. Separate individual cases from systemic problems. One customer’s unusual situation and a repeating pattern call for completely different handling. The first needs individual recovery; the second needs a process change.
3. Set up a mechanism for following up. Reaching out to customers who left a low score and agreed to be contacted is itself an act of service recovery; for the process design, see Complaint Handling and Service Recovery.
4. Report the fix back to the person who raised it. The fact that something changed because of them is, by itself, capable of turning a detractor into a promoter.
5. Review the category counts regularly, to confirm whether the last round of changes reduced feedback in that category.

Three things to settle before you adopt it

NPS adoptions usually fail not because the questions were badly designed but because the preconditions were not in place.

First, confirm somebody will act on the feedback. If there is no established process for contact and improvement once a low score arrives, the survey only creates an expectation and then disappoints it. A customer who was asked and heard nothing back generally thinks worse of you than one who was never asked. For the relevant process, see Complaint Handling and Service Recovery.

Second, confirm the respondent group stays fixed. New customers this time, long-standing customers next time, one specific channel the time after — and the movement in the score becomes uninterpretable. To read a trend, the definition of who is surveyed has to be stable, and any change has to be noted.

Third, confirm the score will not be used as individual performance appraisal. The moment it is tied to frontline staff evaluation, you get steering behavior: asking customers for top marks in person, sending the survey only to people known to be happy. The metric loses its diagnostic value fast, and it usually happens before you notice.

The difference between NPS and a general satisfaction survey is also worth settling early. Satisfaction asks whether this experience was good, which is about the past. NPS asks whether someone is willing to recommend you, which points at future behavior and carries a higher psychological bar because the respondent is putting their own credibility on the line. The two can coexist, but neither should substitute for the other.

One last common misuse: mixing scores collected through different channels, at different moments, from different groups into a single overall number. The figure looks comprehensive and corresponds to nothing you can change. Counting them separately and reading each trend on its own is far more useful than one attractive total.

Compliance boundaries for quoting the score publicly

Before putting NPS into marketing material, a few things have to be settled:

• The score must genuinely exist and be substantiable, including the survey period, who was surveyed, and what was collected
• Do not generalize from a selected sample. A survey sent only to specific satisfied customers is not suitable for claiming as the view of customers overall
• Do not compare against industry averages from unidentified sources; a comparison you cannot substantiate can easily amount to a misleading representation
• Do not package the score as a quality guarantee

For the boundaries around claims and disclosure duties, see Word-of-Mouth Marketing Compliance in Taiwan.

---

The value of NPS is not in the number. It is that the metric forces you to put a very direct question to customers on a regular basis, and to extract a reason from them. What actually changes word of mouth is never the score going up; it is those reasons being taken seriously.

If you want to connect customer feedback to a real word-of-mouth plan, talk to a NETVANA consultant — we help sort out which feedback can become usable external content and which points back at a process that needs fixing first.

Further reading: for a complete performance tracking framework, see How to Measure Word-of-Mouth ROI. For the process that turns a complaint into a recommendation, see Complaint Handling and Service Recovery. For the timing and the hard lines around review requests, see How to Ask Customers for Reviews. And for designing touchpoints as a whole, see The Customer Journey and Word-of-Mouth Touchpoints.