Net Promoter Score (NPS): Calculation, Benchmarks, and Critical Assessment

The Net Promoter Score (NPS) is a metric for measuring customers’ willingness to recommend a company on a scale from 0 to 10. It was developed in 2003 by Fred Reichheld at Bain & Company and published in the Harvard Business Review under the title The One Number You Need to Grow [1]. The central claim: a single question — “How likely are you to recommend us?” — is the best predictor of company growth.

That claim does not hold up empirically. In 2007, Keiningham et al. demonstrated that NPS is no better a growth predictor than other satisfaction and loyalty metrics [2]. Despite this, NPS has established itself as an industry standard — because it is easy to collect, easy to communicate, and hard to ignore.

This article explains how to calculate the NPS, which benchmarks apply across DACH industries, why the “one number” thesis failed, and when NPS is still the right tool — and when you should use CSAT or CES instead.

Where NPS Comes From: Reichheld, Bain, and the “One Number” Thesis

Fred Reichheld, a partner at Bain & Company and author of The Loyalty Effect (1996), sought a metric that would link customer loyalty to company growth. In a study with Satmetrix Systems, he tested 20 different questions on 4,000 customers across 14 industries. His finding: the recommendation question correlated most strongly with actual growth [1].

In 2003, he published the results in the Harvard Business Review. The title — The One Number You Need to Grow — was programmatic: companies should stop running complex satisfaction surveys and instead ask a single question.

In 2006, the book The Ultimate Question followed, and in 2011 the expanded edition The Ultimate Question 2.0, which added the “Closed-Loop Feedback” concept — the systematic follow-up with detractors to understand the root causes of dissatisfaction [3].

The Question

“On a scale of 0 to 10, how likely is it that you would recommend [company/product/service] to a friend or colleague?”

The wording is deliberate. “Recommending” requires more than “being satisfied” — anyone who recommends puts their own reputation on the line. That is why, according to Reichheld, willingness to recommend is a stronger loyalty indicator than satisfaction.

Calculating NPS: The Three Categories

Responses on the 0-10 scale are grouped into three categories:

Category	Score	Behavior
Promoters	9—10	Loyal customers who actively recommend and drive growth
Passives	7—8	Satisfied but unenthusiastic customers — at risk of switching
Detractors	0—6	Dissatisfied customers who spread negative word-of-mouth

The formula:

NPS = % Promoters - % Detractors

Example: 200 customers surveyed. 100 give a 9 or 10 (Promoters = 50%), 60 give a 7 or 8 (Passives = 30%), 40 give 0 through 6 (Detractors = 20%).

NPS = 50% - 20% = +30

The NPS ranges from -100 (all detractors) to +100 (all promoters). A positive NPS is considered good; an NPS above +50 is considered excellent.

What the Category Boundaries Mean — and Why They Are Arbitrary

The boundary between Passives (7—8) and Detractors (0—6) is one of the most debated design decisions in the NPS. Why is a 6 a detractor, but a 7 is not? Reichheld justified the cutoffs with the empirical correlation between ratings and actual recommendation behavior in his studies [1]. Critics such as Keiningham argue that these boundaries vary across cultures and industries — in Japan, for example, a 7 is already strong praise [2].

Practical consequence: A shift of 10 customers from 6 to 7 changes the NPS by 10 points, even though the customer experience has barely changed. This discontinuity makes the NPS susceptible to small sample effects and tactical manipulation.

NPS Benchmarks by Industry in the DACH Region

NPS values are only interpretable in an industry context. An NPS of +30 can be excellent in telecommunications and mediocre in the premium automotive segment.

Industry	Typical NPS Range (DACH)	Context
Automotive	+30 to +50	Premium brands (BMW, Audi, Mercedes) at the upper end; volume brands lower
Insurance	+10 to +25	Low values typical; claims processing as the main driver of detractors
Banking	+20 to +35	Direct banks tend to score higher than branch-based banks
Telecommunications	+15 to +30	Service hotlines as the most common pain point
Energy Utilities	+5 to +20	Low switching propensity distorts the metric’s significance
SaaS / Software	+30 to +50	High variance depending on product category

Important: These ranges are based on industry surveys by Bain & Company, Satmetrix, and NICE (formerly Satmetrix). Exact values vary by methodology, sample, and year. Use them as orientation, not as targets.

Transactional vs. Relational NPS

A frequently overlooked distinction:

Type	Question	Timing	Measures
Relational NPS (rNPS)	“How likely are you to recommend us overall?”	Periodic (quarterly, semi-annually)	Overall relationship with the company
Transactional NPS (tNPS)	“How likely are you to recommend us after this interaction?”	After a specific touchpoint	Quality of a single interaction

When to use which: Relational NPS for strategic management (quarter-over-quarter comparison). Transactional NPS for operational improvement (which touchpoint is generating detractors?). Most DACH companies only collect relational NPS — and miss the operational management information.

The Empirical Critique: Why NPS Is Not the “One” Score

Keiningham et al. (2007): Disproving the “One Number” Thesis

Timothy Keiningham, Lerzan Aksoy, and Bruce Cooil — researchers at Fordham University and Vanderbilt University — replicated Reichheld’s study with a larger sample and more rigorous methodology. Their finding, published in the Journal of Marketing:

“Using longitudinal data, we demonstrate that the Net Promoter metric […] performs no better than other measures of customer satisfaction and loyalty in predicting company growth.” [2]

Specifically, they found:

1. NPS is not a superior growth predictor. The ACSI (American Customer Satisfaction Index) and other satisfaction metrics correlated equally strongly or more strongly with revenue growth.
2. The “one number” claim was never empirically validated. Reichheld’s original study was based on cross-sectional data, not longitudinal data. Correlation is not causation.
3. Different industries require different metrics. The correlation between NPS and growth varies significantly by industry — in some industries, CSAT is a better predictor.

Van Doorn et al. (2013): NPS and Actual Recommendation Behavior

Van Doorn, Leeflang, and Tijs examined the relationship between the intention to recommend (NPS) and actual recommendation behavior. Result: the correlation is weaker than assumed. Customers who give a 9 or 10 do not necessarily recommend actively. And customers who do recommend have not always given high NPS scores [4].

What Remains of NPS?

The critique does not refute NPS as a metric — it refutes NPS as the only metric. NPS is a useful signal, but not the “heartbeat” of the company. Anyone who looks at it in isolation makes worse decisions than someone who triangulates NPS, CSAT, and CES.

Step by Step: Setting Up an NPS Program

Step 1: Define Objective and Scope

Before you send a survey, clarify:

• What do you want to manage? The overall relationship (relational NPS) or specific touchpoints (transactional NPS)?
• Which segments? Private customers, business customers, by contract duration, by product line?
• What decisions should the data inform? If you don’t know the answer, don’t collect the data.

Step 2: Design the Survey

The NPS question alone is not enough. Supplement it with an open follow-up question:

“What is the most important reason for your rating?”

This qualitative question provides the context that the number alone cannot. Without it, you know that you have detractors, but not why.

Survey design rules:

• NPS question first, open question immediately after
• Maximum 3—5 additional questions (survey fatigue sets in from question 6)
• No leading questions (“How satisfied were you with our excellent service?”)
• Mobile-optimized (60%+ of responses come from smartphones)

Step 3: Plan Sample Size and Timing

Sample size: For a statistically reliable result, you need at least 100 responses per segment. At a 20% response rate, that means 500 invitations per segment.

Timing:

• Relational NPS: quarterly or semi-annually
• Transactional NPS: 24—48 hours after the interaction (not immediately, but before the memory fades)

Common mistake: Collecting NPS immediately after a positive experience (e.g., right after contract signing). This creates a positivity bias. Wait 24—48 hours.

Step 4: Analyze the Data

Calculate NPS per segment, not just as an overall score. An overall NPS of +25 can mean:

• Private customers +40, business customers +5 (segment problem)
• New customers +50, existing customers +10 (retention problem)
• Product A +45, Product B -5 (product problem)

Text analysis of open responses: Categorize the free-text responses into a maximum of 8—10 thematic areas. Sort by frequency and segment affiliation. The most frequent themes among detractors are your biggest levers.

Step 5: Establish Closed-Loop Feedback

The most important and most frequently skipped step. Closed-loop means:

1. Inner loop: Within 48 hours, an employee personally contacts every detractor. Goal: understand the cause, solve the problem, repair the relationship.
2. Outer loop: The aggregated insights from inner-loop conversations flow into structural improvements (processes, products, policies).

Without a closed loop, NPS is pointless. You collect feedback but don’t act on it. That is worse than no feedback — because customers notice that their opinion has no consequence.

Practical Example: NPS in Claims Processing at a DACH Insurer

Context: A major DACH insurer has been measuring relational NPS quarterly for two years. The score has stagnated at +18 — below the industry average. The executive board demands “+30 within 18 months.”

Problem: Relational NPS shows that a problem exists, but not where. The team decides to introduce transactional NPS at five touchpoints in addition.

Results after 6 months:

Touchpoint	tNPS	Detractor Share	Top Detractor Theme
Contract signing	+52	12%	Too much paperwork
First invoice	+35	18%	Non-transparent line items
Filing a claim	+8	38%	Unclear process, no status update
Claims processing	-12	52%	Excessive wait time, no dedicated contact
Contract renewal	+28	20%	Price increase without explanation

Insight: Claims processing (tNPS -12) drags down the entire relational NPS. 52% detractors at a single touchpoint. The open responses reveal two dominant themes: (1) customers don’t know where their claim stands, and (2) customers have no dedicated contact person.

Actions taken:

• Automatic claims status updates via SMS/email every 48 hours
• Assignment of a personal claims handler for claims exceeding EUR 2,000
• Service blueprint of the claims process created to identify internal process breakdowns

Result after 12 months: Claims processing tNPS from -12 to +15. Relational NPS from +18 to +27. The +30 target was narrowly missed — but the direction is right, and the organization now understands where to intervene.

Note: This example is illustratively constructed to demonstrate the method in a service context. The structure is based on typical insurance benchmarks.

NPS vs. CSAT vs. CES: When to Use Which Metric?

Three metrics, three perspectives:

Dimension	NPS	CSAT	CES
Question	”How likely are you to recommend us?"	"How satisfied were you with [interaction]?"	"How easy was it to [handle your request]?”
Measures	Willingness to recommend (loyalty proxy)	Satisfaction with a specific interaction	Effort from the customer’s perspective
Scale	0—10	1—5 or 1—7	1—7 or 1—5
Best for	Strategic management, overall relationship	Touchpoint quality, transactional measurement	Service processes, support interactions
Weakness	Not the best growth predictor [2]; culturally biased	Recency bias; says little about loyalty	Only measures effort, not satisfaction or delight
Origin	Reichheld / Bain (2003)	ACSI tradition (Fornell, 1994)	Dixon / CEB (2010)

Recommendation: Use all three — but not everywhere simultaneously. Relational NPS quarterly for strategic management. CSAT after specific touchpoints for operational quality. CES after support interactions and process-heavy services for process optimization.

5 Common NPS Mistakes

1. Treating the Score as a Target Instead of a Diagnostic Tool

Symptom: “Our target is NPS +40” — with no plan for how to improve the customer experience.

Why it hurts: Goodhart’s Law kicks in immediately. Teams optimize the score instead of the customer experience. Typical tactics: surveying only satisfied customers, timing surveys after positive experiences, contacting detractors before the survey.

Solution: Treat NPS as a diagnostic tool, not as a target. The goal is improving the customer experience — NPS is an indicator of whether the improvement is working.

2. Reporting the Score Without Context

Symptom: “Our NPS is +28” — without segmentation, without trends, without comparison.

Why it hurts: +28 can be outstanding (if the industry average is +15) or alarming (if last year’s score was +35). Without context, the number is meaningless.

Solution: Always report NPS with three contexts: (1) trend over time, (2) comparison to industry benchmark, (3) segmentation by customer group and touchpoint.

3. Ignoring Passives

Symptom: All attention goes to promoters (celebrate) and detractors (rescue). The 30—40% passives are overlooked.

Why it hurts: Passives are the biggest churn risk. They are not dissatisfied enough to complain but not enthusiastic enough to stay. A competitor’s offer is enough.

Solution: Analyze what separates passives from promoters. Often a single element is missing (personal contact, proactive communication, one feature) to convert a passive into a promoter.

4. No Closed-Loop Feedback

Symptom: NPS data is collected, loaded into a dashboard, and presented in the quarterly report. Nobody calls detractors.

Why it hurts: Detractors who are contacted after providing feedback frequently become promoters — because the mere fact that someone cares changes the perception. Without a closed loop, you forfeit that potential.

Solution: Establish an inner loop (48-hour callback for detractors) and an outer loop (structural improvements from aggregated insights). Budget 15 minutes per detractor conversation.

5. Using NPS Across Cultures Without Adjustment

Symptom: A DACH company compares its NPS with the US NPS for the same industry — and is alarmed by the difference.

Why it hurts: In German-speaking markets, customers tend to give lower ratings than in the US. An 8 is often strong praise in Germany — in the NPS system, it’s a “passive.” Cultural differences in scale usage make international NPS comparisons unreliable [5].

Solution: Compare NPS only within the same culture and market. Use DACH-specific benchmarks.

When NPS Does NOT Work

1. Monopoly or oligopoly markets: When customers have no switching option (energy utilities with supply obligations, sole provider in a region), NPS measures willingness to recommend that has no consequence. A customer does not recommend their energy provider but stays anyway.

2. Anonymous or one-time transactions: In anonymous e-commerce purchases or one-time services (moving companies, notary appointments), the relationship component that NPS presupposes is missing.

3. Heavily regulated B2B relationships: In industries with long-term framework contracts and high switching costs (IT infrastructure, industrial suppliers), NPS reflects the contract structure, not satisfaction.

4. As the only metric: Never. Keiningham et al. have shown that NPS fails as the sole growth predictor [2]. Combine it with CSAT, CES, and behavioral metrics (churn, repurchase rate, share of wallet).

Frequently Asked Questions

What is a good NPS score?

Industry-dependent. In the DACH region, a rule of thumb: >0 = acceptable, >20 = good, >50 = excellent. But benchmarks vary widely: in the automotive industry, +35 is average; in insurance, +20 is already above average. Always compare NPS within your industry.

How often should NPS be measured?

Relational NPS: quarterly or semi-annually. Transactional NPS: after every relevant interaction (with throttling so customers are not surveyed at every contact — maximum once per quarter per customer).

Is NPS scientifically validated?

NPS is validated as a metric — it reliably measures willingness to recommend. What is not validated is the claim that NPS is the best predictor of company growth. Keiningham et al. (2007) showed that other satisfaction metrics predict equally well or better [2].

What is the difference between NPS and CSAT?

NPS measures general willingness to recommend (loyalty proxy). CSAT measures satisfaction with a specific interaction. NPS is strategic (how is the overall relationship?); CSAT is operational (how was this particular interaction?).

How do you improve NPS?

Not by optimizing the score, but by improving the customer experience at the most critical touchpoints. Step 1: Introduce transactional NPS to identify pain points. Step 2: Analyze open responses from detractors. Step 3: Address the two to three most frequent root causes structurally. Step 4: Establish closed-loop feedback.

Related Methods

A typical service measurement sequence: Use NPS to identify overall loyalty and the most critical touchpoints. Use CSAT to measure satisfaction at specific interaction points. Use CES to find out where the service creates too much effort. Use the Kano model to prioritize which improvements will have the greatest satisfaction impact.

• Customer Satisfaction Score (CSAT): When you want to measure satisfaction at specific touchpoints — more operational than NPS
• Customer Effort Score (CES): When you want to measure effort from the customer’s perspective — especially for support and process-intensive services
• Kano Model: When you want to understand which features drive satisfaction — Kano classifies, NPS measures
• Balanced Scorecard: When you want to embed NPS in a strategic management system — NPS as a metric in the customer perspective
• Measuring Service Innovation: When you want to build a complete measurement framework beyond NPS

---

Research Methodology

This article synthesizes findings from Reichheld’s original publication (2003), the empirical refutation by Keiningham et al. (2007), the analysis of recommendation behavior by Van Doorn et al. (2013), Reichheld’s expanded framework (2011), cultural scale effects in NPS research (De Jong et al. 2015), and DACH-specific industry benchmarks from Bain & Company and Satmetrix.

Limitations: The DACH industry benchmarks are based on aggregated industry surveys with differing methods and samples. Exact values are not directly comparable. The practical example is illustratively constructed, not a documented case study.

Disclosure

SI Labs provides consulting services in the field of service innovation. In service measurement projects, we use NPS as one of several metrics — never as the only one. This practical experience informs the assessment of the method in this article. Readers should be aware of the potential for perspective bias.

References

[1] Reichheld, Fred. “The One Number You Need to Grow.” Harvard Business Review 81, No. 12 (December 2003): 46—54. [Foundational work | Original paper | Citations: 8,000+ | Quality: 80/100]

[2] Keiningham, Timothy L., Bruce Cooil, Tor Wallin Andreassen, and Lerzan Aksoy. “A Longitudinal Examination of Net Promoter and Firm Revenue Growth.” Journal of Marketing 71, No. 3 (July 2007): 39—51. DOI: 10.1509/jmkg.71.3.039 [Journal Article | Empirical refutation | Citations: 1,500+ | Quality: 90/100]

[3] Reichheld, Fred, and Rob Markey. The Ultimate Question 2.0: How Net Promoter Companies Thrive in a Customer-Driven World. Harvard Business Press, 2011. [Book | Expanded framework | Quality: 72/100]

[4] Van Doorn, Jenny, Peter S.H. Leeflang, and Marleen Tijs. “Satisfaction as a Predictor of Future Performance: A Replication.” International Journal of Research in Marketing 30, No. 3 (2013): 314—318. DOI: 10.1016/j.ijresmar.2013.04.002 [Journal Article | Replication study | Quality: 78/100]

[5] De Jong, Ad, Ko de Ruyter, Debbie Isobel Keeling, Stanislav N. Polyakova, and Tom Olsen. “The Interplay Between Net Promoter Score and Cultural Dimensions.” Presented at the EMAC Conference, 2015. [Conference paper | Cultural NPS bias | Quality: 68/100]