What Is NPS (Net Promoter Score)? Calculation, Survey Design, and Score Improvement

Key takeaways:

• NPS measures future advocacy, distinct from CSAT (customer satisfaction), which measures past satisfaction, and CES (customer effort score), which measures how much effort was required. The three metrics are not in conflict—choose between them based on what you measure and at which touchpoint
• The calculation is the percentage of promoters (9–10) minus the percentage of detractors (0–6). Passives (7–8) are not counted, so the same average score can produce very different NPS depending on the distribution
• NPS tends to run structurally low in some markets (notably Japan). It should be read against industry benchmarks and your own time-series change, not by comparing absolute values across countries
• You don't "raise" the score—you move customers through detractor → passive → promoter transitions via closed-loop operations. This guide covers the survey template, the rationale for sample size, and the improvement loop at an implementation level

"Satisfaction is high, yet churn won't stop." "We collect survey scores, but they never lead to the next action." These frustrations are common across customer success and marketing teams. Behind them lies an overlooked truth: satisfaction (the past) and loyalty (future behavior) are different things. NPS (Net Promoter Score) was created precisely to capture that "future advocacy" in a single number.

This article first covers the definition and calculation of NPS with worked examples and a worksheet, then systematically explains the matrix for choosing between NPS, CSAT, and CES; a survey template and the statistical rationale for sample size; why scores run structurally low and how to handle it; closed-loop improvement that moves detractors to promoters; and how to combine NPS with behavioral data. It is a practical guide aimed at the question, "I understand the definition—but how do I actually run it on the ground?"

---

What Is NPS® (Basics and Disambiguation)

NPS® (Net Promoter Score) is a customer-loyalty metric calculated from the distribution of responses to an 11-point scale (0–10) measuring how strongly customers want to recommend a company, product, or service to others. The single question is often described as measuring "net" advocacy.

The key point is that NPS measures not "Were you satisfied?" but the forward-looking intention of "Would you recommend it?" People won't stake their own reputation to recommend something to others unless they genuinely trust the brand or feel attached to it—because if the recommendation disappoints, their own credibility takes a hit. That's why "intention to recommend" represents a deeper commitment than mere "satisfaction," and is considered a stronger predictor of growth-correlated behaviors such as continued use, word of mouth, and additional purchases.

In other words, NPS tries to capture not "the satisfaction a customer states out loud" but "the loyalty a customer is likely to act on." The reason customers churn despite scoring high on satisfaction surveys is that there is a gap between "being satisfied" and "wanting to recommend / keep using." NPS aims squarely at measuring that gap.

Note that the acronym "NPS" varies by context. This article covers the Net Promoter Score of marketing and customer success, but NPS can also mean medical pain scores, a "Network Policy Server" in IT, gaming terminology, and more. Assuming the search intent is business and customer experience (CX), from here on "NPS" consistently refers to Net Promoter Score.

It also helps to recognize that NPS has three facets: a metric, a survey method, and a management system. In the narrow sense it is a "score" (promoters % − detractors %); in practice it can refer to "the survey method used to measure that value"; and in the broadest sense it can mean "the entire management framework for running organization-wide, customer-centric improvement starting from the score." This article first covers NPS as a metric and calculation, then broadens into NPS as a survey method and improvement system. When you say "we're adopting NPS," aligning internally on whether that means simply measuring a score or includes the improvement system will keep discussions on the same page.

On the trademark: NPS® is a registered trademark of Fred Reichheld, Bain & Company, and Satmetrix Systems (now NICE). This article writes "NPS" as a general metric name, but bear in mind it is formally a registered trademark (source: Net Promoter Score - Wikipedia).

The Origins of NPS

NPS was introduced in 2003 by Fred Reichheld of the consulting firm Bain & Company, in the Harvard Business Review article "The One Number You Need to Grow." Reichheld developed the metric in collaboration with Bain & Company and Satmetrix (source: Harvard Business Review / Wikipedia).

Traditional customer satisfaction surveys had many questions, and high scores often failed to translate into business performance or repeat purchases. Reichheld's argument was that the single question "Would you recommend us?" correlates more strongly with company growth than complex satisfaction surveys. This simplicity drove NPS to rapidly become a KPI for companies worldwide. It is also widely used as a core customer success metric.

Part of its spread comes from NPS being easy to adopt as a "common language for management." Because it is a single question summarized into one number, it's easy to share the same goal across departments and to discuss in executive meetings. Marketing, sales, customer success, and product can align around the single point of "creating a state where customers want to recommend us"—and this cross-functional ease is one reason NPS took hold at many companies. At the same time, the very convenience of "boiling it down to one number" is the flip side of the risk that "the score takes on a life of its own," as discussed later.

---

How to Calculate NPS (11-Point Scale, 3 Groups, the Formula)

Calculating NPS simply means dividing respondents into three groups and subtracting the percentage of detractors from the percentage of promoters. Here is the breakdown.

Step 1: Have customers rate on an 11-point scale from 0 to 10

Ask, "How likely are you to recommend this product (service / company) to a friend or colleague?" and have respondents answer on an 11-point scale from 0 (would never recommend) to 10 (would definitely recommend).

Step 2: Classify responses into three groups

• Promoters (9–10): Loyal enthusiasts who love the brand and actively recommend it to others
• Passives (7–8): Satisfied but unenthusiastic; may switch to a competitor
• Detractors (0–6): Dissatisfied customers at risk of spreading negative word of mouth

The crucial point is that 7s and 8s are "passives" and are not counted as promoters. To many people an 8 feels like a "pretty good" rating, but by NPS's design it is not included among promoters. This is one cause of the "structurally low scores" discussed later.

Step 3: Apply the formula

NPS = Percentage of Promoters (%) − Percentage of Detractors (%)

Passives don't appear in the formula. The score ranges from −100 (everyone a detractor) to +100 (everyone a promoter), spanning −100 to +100.

There is intent behind this "ignore the passives" design. NPS's creator believed that passives—satisfied but not actively recommending—contribute little to growth. So the rating is narrowed to the two extremes, "would definitely recommend (promoter)" and "could become a risk (detractor)," and momentum is measured by their difference. It makes sense once you understand it as a way to surface the "imbalance of enthusiasm" that gets buried in an average.

Worked example: Suppose 100 respondents, with 40 promoters, 35 passives, and 25 detractors.

• Percentage of promoters = 40 ÷ 100 = 40%
• Percentage of detractors = 25 ÷ 100 = 25%
• NPS = 40 − 25 = +15

As you can see, no matter how many passives there are, they don't directly affect the score. Understanding that NPS only moves by increasing promoters or reducing detractors pays off in later improvement design.

Score Calculation Worksheet

To picture the actual tally, here is the flow of computing NPS from the number of responses at each point, 0–10. Plug in your own response data and you can get the score with almost no need for a spreadsheet.

In this example, promoters = 10 + 15 = 25%, detractors = 25%, passives = 30 + 20 = 50%. NPS = 25 − 25 = 0. Notice that when responses cluster among passives (7–8), the average looks high yet NPS sinks toward zero. This is a microcosm of the "structurally low scores" discussed later. In practice, you only need to count three figures—promoters, passives, and detractors—and then (promoters − detractors) ÷ total responses × 100 gives the score.

Three Common Misunderstandings in NPS Calculation

Because the formula is simple, the metric is paradoxically prone to misunderstanding. Keep these representative pitfalls in mind.

• It is not an "average score": NPS is not the average of responses. If everyone scores 8 (passive), the average is a high 8.0, yet NPS is 0. Conversely, if half score 10 and half score 0, the average is 5, but NPS is +50 − 50 = 0. NPS's role is to visualize the "split in opinion" that an average hides
• It swings wildly at small scale: With only 20–30 responses, a few ratings move it dramatically. Until you meet the sample size discussed below, don't jump to conclusions of "improvement / decline" based on score fluctuations
• The unit is percentage points, not a percentage: NPS is often expressed as "points," but in substance it is a difference of percentages (percentage points). Reading "NPS = 15" as "15% of people are satisfied" is incorrect

---

NPS vs. CSAT vs. CES (A Matrix for Choosing Between Them)

To use NPS correctly, you need to understand how it differs from the frequently confused CSAT (customer satisfaction) and CES (customer effort score). All three are alike in that they "quantify the voice of the customer," but what they measure, when, and how they ask differ entirely. Many explainer articles treat these three on separate pages, making the distinctions hard to see, so here is one matrix.

Roughly speaking, CSAT measures "Was that experience good? (past)," CES measures "Was it easy? (effort)," and NPS measures "Will you keep doing business and recommend us? (future)." If you want to see the quality of inquiry handling, use CSAT or CES; if you want to track attachment to the brand as an executive metric, use NPS. The roles diverge accordingly.

A concrete example makes the difference vivid. Analyzing churn at one SaaS company, there was a case where "support CSAT was 4.5/5 (high) yet NPS was −20." This means that while individual support was courteous and satisfying, expectations of the product itself and acceptance of the price were low—a state of "not enough to recommend to others." Looking only at CSAT, you'd have misread it as "customers are satisfied." Conversely, leaving a low-CES (cumbersome) touchpoint unaddressed gradually erodes loyalty even when satisfaction is high. A single metric can't capture the customer's state in three dimensions—this is the biggest reason to use the three metrics in combination.

Designing Combined Use (Which to Measure at Which Touchpoint)

In practice, rather than choosing the three metrics exclusively, you place the most suitable one at each touchpoint.

• Right after an inquiry or support is completed → CSAT / CES (evaluation of that response itself)
• At onboarding completion or cancellation → CES (detect friction in the process)
• After a set period from contract / annually → NPS (periodic monitoring of overall loyalty)

Designed this way, you can monitor both "the quality of individual experiences (CSAT/CES)" and "the health of the overall relationship (NPS)." Because a proliferation of metrics increases respondent burden, the operational tip is to narrow to one based on the importance of the touchpoint. Connecting this thinking to health scores and customer success KPI design, which measure customer state alongside NPS, improves visibility across your metric set.

---

Designing an NPS Survey (Template and Required Sample Size)

NPS isn't "done once you run the survey." The skill of survey design—which customers, at what timing, asked what—determines the reliability of the resulting score and its connection to improvement.

Relational NPS vs. Transactional NPS

NPS surveys split into two types based on purpose.

It helps to think of relational NPS as a "full-body checkup of the whole company," and transactional NPS as a "detailed examination of a specific touchpoint (onboarding, support, pre-renewal, etc.)." Combining the two lets you visualize both the overall level and bottleneck touchpoints.

In practice, a sequence that works well is: first grasp the overall score and rough problem areas with relational NPS, then deploy transactional NPS at touchpoints suspected of causing low scores to dig deeper. For example, if a relational survey reveals "early-stage customers score low," you can place a transactional NPS at onboarding completion to pinpoint which step harms the experience. Relational alone won't tell you "where it's bad," and transactional alone won't show "how things are overall." Going back and forth between checkup and detailed exam raises the precision of improvement.

NPS Survey Template

A survey is fundamentally a three-part structure: "the core advocacy question" + "an open-text field asking for the reason" + "attribute questions for analysis." Here it is in a ready-to-use form, without relying on a separate download.

[NPS Survey Template (Relational version)]

Q1. (Core question, required)
How likely are you to recommend "○○ (your service name)" to a
friend or colleague?
Please answer from 0 (would never recommend) to 10 (would
definitely recommend).

  0  1  2  3  4  5  6  7  8  9  10

Q2. (Reason, open text / the most important question, the
     starting point for improvement)
Please describe, to the extent you're comfortable, the specific
reason for the score you gave.
(e.g., what was good / what you'd like improved)

Q3. (Attributes, optional / for analysis segments)
- Length of use (< 6 months / 6 months–1 year / 1–3 years / 3+ years)
- Primary department / role using the service
- Contract plan

──────────────────────────────
[Differences for the Transactional version]
Tie the wording of Q1 to the touchpoint, e.g.:
"Based on your recent support experience, how likely are you to
 recommend ○○ to a friend or colleague?"
Add to Q3: "the relevant touchpoint (support / onboarding, etc.)"

The most important item is Q2, the open-text field. The score is merely "a number for the present"; the hints for improvement always reside in the reason comments (VoC = Voice of Customer). A design that collects only scores without asking why cannot run the later improvement loop.

How to Think About Required Sample Size

"How many people need to answer?" is a question that always comes up in practice. Statistically, when the population is large enough, the required sample size is given by the following formula.

n = z² × p(1−p) ÷ e²

• z = the value for the confidence level (1.96 for 95%)
• p = the population proportion of responses (use the most conservative 0.5 if unknown)
• e = the margin of error (0.05 for ±5%)

Plugging in: n = 1.96² × 0.5 × 0.5 ÷ 0.05² = 0.9604 ÷ 0.0025 ≈ 385.

In other words, with a ±5% margin of error and 95% confidence, roughly 385 valid responses (400 as a practical target) lets you estimate the overall population trend with a certain precision. For smaller mid-size companies with fewer customers, the population correction reduces the required number further. The key is not to vaguely collect "the more the better," but to work backward from the precision needed for the decision. If you want to analyze by segment (by plan, by tenure), design so each segment meets this level.

Response rate can't be ignored either. In B2B relational surveys, response rates of 10–30% are not unusual, so if you want 400 valid responses, you need to plan for several times that many invitations. To raise the response rate, narrow the number of questions (keep it to about three: the core NPS question + reason + attributes), state the time required, and explain that responses will be used for improvement. Also watch for "non-response bias," where respondents and non-respondents differ in tendency. When only highly engaged customers answer, the score skews optimistic relative to reality, so interpret scores cautiously when the response rate is extremely low.

---

NPS Averages and the Yardsticks for a "Good Score" (Handling Market Bias)

"Our NPS was +15—is that good or bad?" This is the most common NPS question. The conclusion: you can't judge good or bad from the absolute NPS value alone, because the standard differs completely by industry, country, and survey type.

In some markets—Japan being a notable example—NPS tends to run structurally low (negative). In industry benchmark surveys by NTT Com Online (now NTT DOCOMO Business X), many industries show negative averages. For instance, the average across 11 automotive brands was −22.8 (with even the top, LEXUS, at +17.4), and the average across 5 security-software companies was −32.0 (with the top, ESET, at −16.5) (source: NTT DOCOMO Business X "NPS Industry Benchmark"). EmotionTech also notes that "−15 or below is common" in the Japanese market (source: EmotionTech "Industry Average NPS Scores").

For reference, here are published industry NPS figures reported in Japan (values vary by survey organization, year, and number of companies, so treat them only as a sense of the level showing that "negative territory is common").

(Source: NTT DOCOMO Business X "NPS Industry Benchmark Rankings", from each industry's published figures. Automotive and security software are from the 2024 survey; face-to-face brokerage, electric power, and credit cards are from each industry's most recent survey. Figures vary by survey year.)

In other words, in Japan "negative NPS" doesn't necessarily mean "bad." If the same industry's average is −30, then −15 is relatively good. Applying overseas standards like "+30 is excellent" directly leads to misplaced alarm or complacency. Keeping in mind that industry averages tend to land in positive territory only in a few high-loyalty industries (luxury cars, some SaaS, etc.) will keep you from misjudging.

Why NPS Runs Low in Some Markets

Behind low scores in markets like Japan are cultural response tendencies.

• Central tendency: Respondents avoid the extremes (0 or 10) and cluster in the middle (7–8). Because 7–8 count as passives in NPS, promoters don't grow and the score is slow to rise
• High-context culture: It's hard to express "satisfaction" through explicitly high ratings, so people give modest scores

This doesn't mean service quality is low. Because the measuring stick itself differs by country, be careful with international comparisons of scores.

Three-Lens Evaluation on the Premise of Low Scores

If absolute comparison is difficult, read the score through these three lenses.

1. Peer-relative: Compare against benchmarks of the same industry and the same survey method. The gap from the industry average is your true standing
2. Own time series (Δ): Track change from your own previous score. From −20 to −12 is clear improvement, even if the absolute value is negative
3. By touchpoint: Use transactional NPS to break it down by touchpoint and pinpoint which experience is dragging you down

Of these three, what teams should prioritize most is the own time series (Δ). Because industry benchmarks don't perfectly match your survey method and target, strict comparison is hard—whereas comparison with your own prior score keeps conditions aligned and most honestly reflects the effect of your improvements. The fact that "we improved 8 points, from −20 to −12" is solid evidence your measures are working, even if the absolute value is negative. Conversely, even if the absolute value exceeds the industry average, a downward trend is a warning sign.

NPS is essentially used not to declare "we beat / lost to a competitor" but as a yardstick for continuous self-improvement. Tenaciously tracking your own change yields far more insight than agonizing over comparisons with others.

---

Don't Raise the Score—Move the Segments: Closed-Loop Improvement

The most common failure in NPS operation is making "raising the score" itself the goal. The score is a result, not something to manipulate. Correctly, you design interventions that move detractors to passives and passives to promoters. The mechanism that creates these transitions—never leaving feedback unattended, always connecting it to action—is the closed loop.

A closed loop is easier to operate when broken into three loops with different targets and time horizons.

① 48-Hour Follow-up with Detractors

Left alone, detractors lead to negative word of mouth and churn—but if you engage promptly, they are also the segment with the greatest room for loyalty recovery. As soon as possible after a response (within ~48 hours), a representative contacts them directly, confirms the nature of the dissatisfaction, and commits to a concrete response. The very experience of "they responded right away" lifts the rating.

What matters here is placing the purpose of follow-up on "listening and correcting," not "defending." Denying or making excuses for the low score only deepens dissatisfaction. First listen fully, confirm the facts, candidly convey what you can and cannot do, and commit to a concrete next action with a deadline. When this sequence is handled sincerely, a detractor can come to feel "I trust you more than before"—sometimes even turning into a promoter. Given that customers leave less over the dissatisfaction itself than over a poor response to it, detractor follow-up is the front line of churn prevention. Recording who responded by when, and visualizing the follow-up rate as a KPI, turns this from individual heroics into an organizational system.

② Root-Cause Fixes via VoC Tag Aggregation

Individual follow-up alone keeps producing the next detractor from the same dissatisfaction. Classify and aggregate the Q2 open text (VoC) into tags like "price," "support quality," and "missing features," and fix the product or operations starting from issues high in frequency × impact. Monitoring whether the NPS tied to that issue rises afterward lets you verify the effect of your measures.

Note here that "issues with many mentions" and "issues with large impact on the score" don't necessarily coincide. There are cases where comment counts are low but the customers who mentioned that issue are uniformly detractors (= large contribution to the score). Conversely, there are issues with many voices but trivial requests, where resolving them barely moves the score. So by also looking at "the average score of customers who mentioned it" and "the detractor rate" for each issue tag, you can quantitatively see where a fix would move NPS the most. Because improvement resources are finite, this prioritization determines return on investment. After implementing an improvement, always check "did that issue tag's score improve in the next survey?" and reconsider measures that had no effect—pairing this verification ensures VoC analysis doesn't end at "aggregate and done."

③ Move Promoters to Referrals and Reviews

Promoters merely "want to recommend"; without an actual chance to do so, it doesn't translate into business contribution. Provide promoters with "places to recommend" such as referral programs, case-study interviews, and posting to review sites. The voice of promoters becomes a powerful asset for new acquisition, and directly drives revenue expansion (NRR improvement) through upselling and cross-selling to existing customers.

One thing easily overlooked among the three loops is engaging the passives (7–8). Passives lack the urgency of detractors and the visibility of promoters, so they tend to be neglected—yet they are often the largest segment by count. Because NPS also moves by increasing promoters, "passives who just need one more push to become promoters" are the highest-leverage target. Passives' open text often contains concrete improvement requests like "mostly satisfied, but if only ○○," and if you pick these up and deliver an experience that exceeds a small expectation, a transition to promoter occurs. Not overlooking passives while busy putting out detractor fires is the trick to continuously lifting the score.

Rather than trying to "raise" the score, the score moves as a result of continuously running these three loops—not mistaking this order is what separates success from failure in NPS operation.

---

Five Steps to a Successful NPS Rollout

Let's rearrange everything so far into a chronological procedure for actually launching NPS internally. Many rollout failures stem from the wrong order: "blast it company-wide all at once, then get stuck on what to do with the score once it appears." Proceeding through the following five steps lets you design end-to-end through to improvement.

Step 1: Narrow to one purpose and one touchpoint

The cardinal rule is not to deploy to all customers and all touchpoints from the start. Decide one purpose—"improve the post-onboarding experience," "report annual overall loyalty to management"—and choose the survey type (transactional / relational) and touchpoint that fit it. A score taken with an ambiguous purpose will always leave you unsure how to use it.

Step 2: Design the survey and distribution

Based on the template above, prepare the core question, the open-text reason, and attribute questions. Decide the distribution channel (email, in-app, web, etc.) and the invitation count and collection target back-calculated from the required sample size. For transactional surveys, also design "within how many hours of the touchpoint to distribute."

Step 3: Calculate the score and break it down by segment

Calculate NPS from the responses collected, and break it down not just into the overall score but by plan, by tenure, by touchpoint. Even if the overall is −15, viewing by segment—"customers under one year are at −40"—instantly clarifies where the problem lies. Speaking with distribution and breakdown rather than averages is the basis of analysis.

Step 4: Aggregate VoC into issue tags

Classify open-text comments into tags like "price," "features," "support," and "ease of use," and rank them by frequency and impact. This turns "the reason the score is low" from a qualitative impression into a prioritized list of improvement themes.

Step 5: Run the closed loop and verify the effect

Execute the three loops from the previous chapter (immediate follow-up, root-cause fix, promoter activation). After implementing an improvement, always check in the next survey whether the score tied to that issue tag has risen. By "not firing and forgetting," NPS transforms from a mere periodic observation into an engine that drives an improvement PDCA cycle.

These five steps aren't a one-and-done; repeat them on a quarterly or annual cycle. With each round, you accumulate a sense of what a "good score" means for your company and patterns of which improvements work.

---

NPS × Behavioral Data: Combining Qualitative and Quantitative with a DSR

NPS has a structural weakness: it can only capture a "qualitative point" at one moment when the customer answered the survey. Respondents are only a fraction of all customers, and you can't notice when a customer's heart drifts away between responses. Noticing only after the score has dropped is too late for detecting churn signals.

Only a fraction of customers answer surveys in the first place, and the true feelings of the silent majority remain invisible. What fills this weakness is combining it with the customer's behavioral data (quantitative and continuous). In recent years, using a Digital Sales Room (DSR)—an online space consolidating deal materials, proposals, and adoption guides—you can continuously visualize who viewed what, and how much. Unlike the internal deal data managed by SFA or CRM, this is a quantitative signal that reflects the movement of the customer's own interest.

Combining NPS (qualitative) with viewing engagement (quantitative) on two axes lets you divide customers into the following four segments and vary your approach.

For example, a customer who "has low NPS but frequently views guides and update notices" is a recoverable segment that still retains interest despite dissatisfaction. Conversely, a customer whose "NPS and viewing are both declining" should be addressed early as the top-priority churn-signal target. Meanwhile, a customer with "high NPS and active viewing" is at the perfect moment to be offered a referral or upsell/cross-sell.

The value of combining with behavioral data is that you can bring forward customer responses that tend to lag when you "act only on the NPS score" by using quantitative signals. Embedding this data into inside sales and customer success workflows enables reproducible customer handling that doesn't rely on intuition or memory. In an actual case where a CS team used a DSR as well, visualizing customer interest became the starting point for early follow-up.

This idea can be framed as filling NPS's "can only be measured at a point (the survey moment)" limitation with behavioral data's "line (day-to-day engagement)." NPS tells you "why they feel that way (the reason = VoC)," while behavioral data tells you "how they're actually moving right now (engagement strength)." The former indicates the direction of measures; the latter indicates the timing of action. When you can cross-reference both on the same customer ledger, the field dilemma of "I know the reason for the dissatisfaction, but I don't know when to reach out" is resolved.

For example, suppose a customer who was a detractor in the quarterly NPS survey then starts eagerly viewing a new adoption guide in the DSR. This is a sign of "dissatisfaction remains, but they're still trying to make it work," and if CS brings a proposal at exactly this timing, it's easier to trigger a detractor → passive transition. Acting on a trigger of behavioral change rather than waiting for score aggregation—this operation can never be achieved by the qualitative NPS metric alone.

---

eNPS and How to Overcome NPS's Drawbacks

What Is eNPS (Employee NPS)?

Applying NPS's thinking to employees gives eNPS (Employee Net Promoter Score). It asks "how likely are you to recommend this company as a place to work to a friend or acquaintance?" on a 0–10 scale, and scores it with the same formula. In contrast to customer NPS, which measures customer loyalty, eNPS is used as a metric for employee engagement, for organizational improvement and turnover prevention.

The value of deliberately viewing customer NPS and eNPS side by side lies in the relationship that "customers rarely want to recommend the service of a company whose own employees wouldn't recommend it." Customer experience (CX) is ultimately delivered through the experience (EX) of employees—especially the sales, support, and CS who stand at customer touchpoints. When frontline employees are worn down and can't take pride in their own service, courteous customer handling and improvement proposals won't last. So the root cause of stagnating customer NPS sometimes actually lies in declining employee engagement. The growing number of companies tracking CX and EX as a set is because they've noticed this causality. As with customer NPS, operating eNPS isn't just about "taking a score"—the key is to collect reasons via open text and connect them to fixing organizational issues.

Answering Why People Say "NPS Is Meaningless"

NPS also faces criticism that it's "meaningless" or "useless." Here are the main points and how to engage with them.

• Limits of a single metric: One question can't capture all of a customer's complex psychology. → It can be complemented by using it in combination with CSAT, CES, churn rate, and behavioral data
• Sample bias: Respondents skew toward the highly engaged. → Mitigate this through distribution-target design, response-rate management, and securing the required sample size
• The harm of score supremacy: Chasing only the number leads the field to coach responses or hollow out improvement. → Set VoC and transitions (improvement actions), not the score, as the KPI

Most of these criticisms stem from "using NPS in isolation, chasing only the score." As discussed throughout this article, on the premise of combining it with other metrics, leveraging open text, and running closed-loop operations, NPS remains a powerful compass. Operational design, more than the metric itself, determines success.

A practical landing point is "not making NPS the only North Star." Report NPS to management as a summary metric showing the overall trend of loyalty, while running daily improvement on finer-grained metrics like CSAT, CES, churn rate, and behavioral data. NPS is, after all, a compass for confirming "are we heading in the right direction?"—not the steering wheel itself. If you can share this positioning internally, you can avoid falling into "surveys for the sake of the score."

---

How to Choose NPS Operation Tools

NPS surveys can be started even in a spreadsheet, but considering continuous operation, automated analysis, and the closed loop, using a dedicated tool is realistic. When selecting, compare on the following points.

• Distribution and collection automation: Channel-appropriate distribution (email, web, in-app) and automatic triggers for transactional surveys
• Open-text (VoC) analysis: Whether text mining or classification can aggregate comments into issue tags
• Segment analysis: Whether scores can be broken down by plan, tenure, and touchpoint
• Integration with existing systems: Whether it integrates with CRM/SFA or DSR behavioral data to view qualitative × quantitative together
• Closed-loop support: Whether follow-up operations—detractor alerts, task creation—can be built into a system

A tool is, after all, a means to support operations. Only when the design of "which score, at which touchpoint, used by whom for improvement" comes first does a tool come alive. Even with a high-function tool, if the purpose and operational flow aren't set, it ends up as just "a dashboard that displays scores."

A realistic approach scaled to your size is to start small with just one touchpoint using a spreadsheet and a survey form, then migrate to a dedicated tool once operations are running. Rather than building an elaborate system from the start, creating a "small success with one focused purpose" and rolling it out laterally is easier to gain internal buy-in for and to make stick. Especially in B2B, where the number of customers is limited, deeply reading each company's VoC works well—so it's worth spending time on the design of "how to turn the voices you pick up into improvement" rather than on automation.

---

Conclusion

Here are the key points for running NPS in practice starting tomorrow.

• NPS measures "future advocacy." Distinguish its role from CSAT (which measures past satisfaction) and CES (which measures effort), and use them by touchpoint. The metrics aren't in conflict—they're used together
• The calculation is promoters % − detractors %. Because passives (7–8) aren't counted, understanding the score distribution is a prerequisite for improvement design
• Scores run structurally low in some markets. Don't agonize over absolute values; read them through the three lenses of peer benchmark, own time series, and by-touchpoint
• Don't raise the score—move the segments. Keep running the three closed loops: 48-hour follow-up with detractors, root-cause fixes via VoC tag aggregation, and promoter referral/review activation
• Fill the weakness of NPS (qualitative, a point) with behavioral data (quantitative, continuous). Combine it with DSR viewing engagement to bring forward churn-signal detection and promoter-candidate identification

NPS changes nothing if you only take it. Only by continuously running the loop of "measure → ask why → create transitions → verify the effect," combined with customer behavioral data, does it become a compass that leads to business growth. Rather than fretting over the ups and downs of the score, the royal road to long-term loyalty improvement is the posture of looking at the customer voices and behaviors behind it and accumulating improvements. Start by designing one NPS survey with a focused purpose at one of your key touchpoints. The experience of starting small and running improvement is the first step toward rooting a "customer-centric culture" in your organization.

※ Sources for this article:

• NPS origin, creator, and trademark: Net Promoter Score - Wikipedia (Fred Reichheld, "The One Number You Need to Grow," HBR, 2003)
• Japan industry NPS benchmarks: NTT DOCOMO Business X "NPS Industry Benchmark Survey" (2024 survey)
• Japanese market score tendency: EmotionTech "Industry Average NPS Scores"