NPS Is Theater: Why Boards Love It and Customers Don't Care

Q: Does NPS actually predict company growth?

The evidence is weak. The landmark 2007 Journal of Marketing study by Keiningham and colleagues found no support for the claim that NPS is the single most reliable indicator of a company's ability to grow, and failed to replicate its clear superiority over other metrics. A balanced MeasuringU assessment graded the growth-prediction claim a C+ — useful, but nowhere near the one number you need to grow.

TL;DR

Board-deck NPS is the most-tracked metric in SaaS and the least-acted-on one. The single number on slide 14 survives because it's legible to a board, not because it tells your product team what to ship. Track the free-text, not the score.

NPS is theater when it lives on a quarterly slide and dies there. The number goes up, the number goes down, the room nods, and nobody can name the specific change they'll make because of the move from 31 to 34. That's a ritual that produces a slide, not a measurement.

The practitioners closest to the work already feel it. Gartner predicted in 2021 that more than 75% of organizations would stop using NPS to measure customer service and support success by 2025. It didn't die, but it slid, dropping from the second-most-popular CX metric in 2023 to eighth in 2024, per CMSWire's 2025 report, and the teams still leaning hardest on it are the least mature (43% of less-mature teams use NPS versus 30% of teams with stronger systems). The underlying data is fine; the ritual built on top of it, a company-wide aggregate surveyed inconsistently and acted on by nobody, is the problem. This post is a takedown of the ritual, not the data.

This post covers:

How NPS became a board ritual (Reichheld, HBR 2003, "one number")
The four structural problems: sample bias, cross-cultural skew, the score-vs-action gap, and category manipulation
What leadership actually wants NPS to do, and why the ritual version can't do it
The replacement practice: track the free-text, segment by journey stage, pair with behavior, interview churners
A table comparing NPS-as-ritual to NPS-used-diagnostically, and the honest caveat

How Did NPS Become a Board Ritual?

NPS became a board ritual because it compressed messy customer sentiment into one number a non-specialist executive could read in three seconds, and that number was sold as the single best predictor of growth. Fred Reichheld, a partner at Bain & Company, introduced it in a 2003 Harvard Business Review article, "The One Number You Need to Grow". The pitch was in the title. One number. That's the part the board kept.

A board doesn't want a 40-page sentiment analysis; it wants a trend line, a number with an arrow, and a benchmark, and NPS delivers all three on one slide. The metric won the leadership layer not because it was the most diagnostic tool available, but because it was the most legible one. Legibility and diagnostic value are different properties, and the gap between them is where this whole post lives.

By 2026, NPS is still near-universal in B2B SaaS: surveyed on a 0-10 scale, reported as a company-wide aggregate, reviewed quarterly. Questioning it feels like questioning whether you should measure customers at all. You should. The question is whether the quarterly aggregate on the board deck is the measurement, or just a costume worn by one. For how NPS stacks up against the other survey metrics, our NPS vs CSAT vs CES comparison covers the mechanics before we take it apart.

Why Is Board-Deck NPS the Most-Tracked Metric That Nobody Acts On?

Board-deck NPS goes unacted-on because the company-wide aggregate strips out everything you'd need to act. A single number for the entire customer base tells you sentiment moved. It doesn't tell you which segment moved, why it moved, or what to ship in response. The number is a thermometer, and you can't fix a fever by staring at a thermometer.

Here's the test. If your NPS moved from 31 to 34 last quarter, can your product team name the specific feature, fix, or workflow change they'll ship because of it? Usually the honest answer is no. The number changed, the slide got updated, and the roadmap stayed exactly where it was. If you can't name the specific change you'll ship based on a 7-to-8 NPS move, the metric is decoration. The actionable signal isn't in the score; it's in the open-text comment next to it, which is the part most board reviews skip. As product-strategy expert Itamar Gilad, a former Google and Microsoft product lead, notes, that open follow-up ("What's the main reason for your score?") is "the least contentious and the most valuable" part of the survey. The board gets the number; the comments, where the product intelligence lives, get dropped into a word cloud.

That's why NPS is simultaneously the most-tracked and least-acted-on metric in the category. Tracking the number is cheap and legible; acting requires the messy half nobody put on the slide, which is the exact gap the qualitative-versus-quantitative feedback distinction is built around.

What Are the Four Structural Problems With NPS as a Ritual?

NPS-as-ritual fails on four independent axes: who actually responds (sample bias), how culture distorts the scale (cross-cultural skew), what the score can't tell you (the score-vs-action gap), and how easily targets corrupt it (category manipulation). Each is enough to undermine a board's confidence on its own. Together they make the quarterly aggregate close to meaningless as a research input, even though the same data, used differently, would be fine.

Problem 1: Sample Bias — Only the Loudest Reply

NPS over-represents the extremes because the people who bother to answer a survey are systematically different from the people who don't. The satisfied and the furious respond; the quietly-indifferent middle ignores it. Your "company-wide" number is really a number from the minority who self-selected into answering, and that minority skews toward strong opinions.

The math makes this concrete. The average B2B NPS survey response rate is around 13%, and frustrated customers are the least likely to respond — Gilad notes UX researcher Jared Spool puts it as low as 4-7%. Even a "good" rate doesn't rescue the sample: B2B CX consultancy Genroe considers anything above roughly 20% a good NPS response rate, with well-crafted surveys landing between 10% and 30%. A good result still leaves 70-80% of your customers unheard, and the silence skews unhappy, because the customer already halfway out the door is exactly the one who won't spend 20 seconds rating you. For why the quietest customers carry the most diagnostic weight, see how to get user feedback when nobody responds.

Problem 2: Cross-Cultural Skew — A "9" Means Different Things in Different Countries

NPS is nearly meaningless across markets because the 0-10 scale is read through cultural norms about extreme ratings, and those norms vary enormously. Americans rate high; many European and East Asian customers rate conservatively, not because they're less satisfied, but because a perfect score violates a cultural norm. A global NPS aggregate averages incompatible scales.

The spread in real data is staggering. SurveyMonkey's cross-country benchmark puts Brazil at 62, India at 51, the US and Australia at 30, the UK at 28, the Netherlands at 2, and Japan at -52 — a 114-point gap in one study. Peer-reviewed research explains the mechanism: a study in Survey Practice by Wang, Hempton, Dugan, and Komives found Asian and Asian American respondents chose the middle option substantially more often, and extreme options less often, than White respondents, significant on 53 of 68 items (78%). Since NPS only counts 9s and 10s as promoters, a culture that avoids the top of the scale posts a structurally lower score regardless of loyalty. A board comparing this quarter's global NPS to last quarter's, when the response mix shifted between cultures, is reading noise as signal. (For why even same-country benchmarks need context, see average NPS scores by SaaS category.)

Problem 3: The Score-vs-Action Gap — A "9" and a "10" Tell You Nothing About What to Fix

The score contains no instructions. Knowing a customer rated you 9 instead of 10, or 6 instead of 8, tells you the direction of their sentiment and nothing about its cause. You can't build a roadmap from a distribution of integers. The decision-relevant information is why they chose the number, and the number doesn't carry the why.

The sharpest critics have hammered this for years. Jared Spool, co-founder of the UX school Center Centre and founder of User Interface Engineering, wrote in his 2017 essay "Net Promoter Score Considered Harmful" that NPS scores "are the equivalent of a daily horoscope. There's no science here, just faith," concluding flatly, "We consider NPS to be harmful." Even balanced assessments agree the score is weak alone: Jeff Sauro, PhD, and Jim Lewis, PhD, of the research firm MeasuringU graded the claim that NPS is "predictive of future company growth" a C+ — nowhere near the "single most reliable indicator" the ritual treats it as. Gilad's practical rule follows: don't use NPS "as a key result in your OKRs, as an A/B experiment metric, or try to directly influence it — too many factors go into user satisfaction and it's not actionable on its own". The action lives one column over, in the free-text, which is why the replacement practice starts there.

Problem 4: Category Manipulation — Hitting the Target by Choosing Who You Ask

NPS is trivially gameable because the score depends entirely on who you survey and when, and both are usually controlled by the team being measured. Survey only your happiest cohort, time it right after a positive interaction, exclude the accounts about to churn, and the number climbs without a single thing improving for customers.

The most damning witness is the creator of NPS. In the 2021 follow-up "Net Promoter 3.0," Reichheld and co-authors Darci Darnell and Maureen Burns acknowledged how widely the metric is misused — above all when companies tie NPS to frontline evaluations and compensation, which invites what they characterized as "pleading, bribery and manipulation". The tactics are mundane and widespread — CX firm Satrix Solutions catalogs begging for high scores, "leading the witness" with messaging that a 9 or 10 is expected, timing surveys for the post-event halo effect, hand-picking respondents while excluding likely churners, and tying bonuses to the score. This is Goodhart's Law in a board deck: when a measure becomes a target, it stops being a good measure. A ritual you can satisfy by changing who you ask isn't measuring your customers; it's measuring your survey design.

What Does Leadership Actually Want NPS to Do?

Leadership wants NPS to be a single, trustworthy gauge of product health the board can track over time, the way a CFO tracks gross margin. That's a reasonable want. The problem is that the company-wide quarterly aggregate, built on a self-selected sample and a culturally-skewed scale, can't be that gauge, because the operation that makes NPS legible (compressing customer health to one number) is the exact operation that destroys the information needed to manage it.

The academic record is blunt about it. The landmark study is Keiningham, Cooil, Andreassen, and Aksoy's 2007 paper in the Journal of Marketing, "A Longitudinal Examination of Net Promoter and Firm Revenue Growth." Using data from 21 firms and over 15,500 interviews, they set out to replicate Reichheld's claims and, in their own words, found "no support for the claim that Net Promoter is the 'single most reliable indicator of a company's ability to grow.'" The paper won the Marketing Science Institute's H. Paul Root Award, and the Ehrenberg-Bass Institute has since amplified the critique to the point of calling NPS's growth claim "fake science." Read the evidence generously and it still lands short: in the MeasuringU report card, NPS earned an A for reliability but only a C+ for predicting growth. The thing the board wants NPS to be — the one number that predicts growth — is the specific claim the research declines to support. That's the throughline of using feedback to reduce churn instead of watching a score.

What Should You Track Instead of Board-Deck NPS?

Replace the quarterly company-wide score with four practices that capture what the ritual throws away. None require killing your NPS survey; they require using the data it generates like a research instrument instead of a slide. If you do nothing else after reading this, install these four. They cost almost nothing beyond attention.

1. Track the Free-Text, Not the Score

The single highest-leverage change is to read and theme the open-text comments and report those to the board, with the score as a footnote. The comment is where the customer tells you what to fix; the number is where they tell you how they feel about it, which you mostly already know. Clustering a few hundred verbatims a quarter by hand is where this plan usually dies, which is the exact job AI feedback-analysis tooling exists to do — it's the layer FeedSense sits in — but the discipline matters more than the tool. A board that hears "detractors are clustering around a three-step export workflow that takes 90 minutes" can make a decision; a board that hears "NPS is 34, up two points" cannot. Turning raw comments into themes is the whole subject of the complete guide to feedback analysis.

2. Segment by Journey Stage, Not Company-Wide

Stop reporting one number for the entire base. A new customer in week two and a power user in year three are answering completely different questions, and averaging them produces a number that describes nobody. A company-wide 34 might hide an onboarding cohort at 5 and a mature cohort at 60 — the aggregate says "fine," the segmentation says "your activation experience is bleeding detractors." Same responses, radically different decision. If you must compare markets, compare each to its own trend line, never to a global average, because a 9 in the US and a 9 in Japan are not the same statement.

3. Pair NPS With a Behavioral Metric

Always read NPS next to what customers actually do, because stated sentiment and revealed behavior diverge constantly. A customer who rates you 9 but hasn't logged in for three weeks is telling you the truth in their behavior and a courtesy in their survey. Pair the score with retention at N days, feature adoption, or usage frequency, and let the behavioral metric arbitrate when the two disagree. Behavior is harder to game than a survey; A/B testing expert Ronny Kohavi, quoted in Gilad's analysis, argues behavior metrics automatically collected in experiments are likely better than periodically surveying large populations. A score floating free of behavioral data is the most manipulable version of the metric.

4. Run Qualitative Exit Interviews With Churners — Ten Beats a Hundred

For the customers who matter most to your roadmap, skip the survey and call them. Ten 30-minute interviews with recently-churned customers produce more usable insight than a hundred NPS responses, because churners have no reason left to be polite and every reason to tell you exactly what broke. This is the inverse of the sample-bias problem: NPS excludes the silent majority, while churner interviews go straight to the people who left and ask them why, in their own words. The mechanics of feeding that signal back into the product are covered in how to close the feedback loop. Ten honest conversations with people who fired you will reshape a roadmap faster than any aggregate score ever has.

NPS as Ritual vs. NPS Used Diagnostically: The Side-by-Side

The honest comparison isn't "NPS bad, something else good." It's that the same NPS data behaves completely differently depending on whether you treat it as a board ornament or a research input. The ritual column kills the metric's usefulness; the diagnostic column is the same survey used like an instrument.

Dimension	NPS as Ritual	NPS Used Diagnostically
What gets reported	A single company-wide number	Themed open-text + score as footnote
Unit of analysis	Entire customer base, averaged	Segmented by journey stage
What the number drives	A slide, a trend line	A specific, named roadmap change
Free-text comments	Aggregated to a word cloud or dropped	Read, themed, quoted — the headline
Sample	Whoever self-selected (~13% in B2B)	Score read against behavior + churner interviews
Cross-market comparison	Global average across cultures	Each market vs. its own trend
Gameability	High (survey only happy cohorts)	Low (anchored to behavior customers can't fake)
Cadence	Quarterly, for the board	Continuous, for the product team
Tied to comp/OKRs	Often (creates inflation incentive)	No — a diagnostic, not a target
Predictive of growth	Treated as "the one number"	Treated as one weak signal among several

This isn't a takedown table; it's a routing table. The left column is the costume, the right column is the same metric doing actual work, and the difference in cost between them is essentially zero — you're already collecting the data. The wrong move is reporting the left column and calling it customer research.

Isn't NPS Useless, Then? The Honest Caveat

No, and the post falls apart the moment you read it that way. NPS is not useless. It's misused. A representative sample, read alongside the open-text comment and a behavioral metric, segmented by journey stage and tracked against its own history rather than a cross-cultural average, is a genuinely useful gauge of sentiment over time. As a single team's trend line watched for direction rather than absolute value, it's a reasonable early-warning indicator — the same MeasuringU report card that graded the growth claim a C+ gave an F to the idea that NPS should never be used. The metric earns a real, if modest, place. What it doesn't earn is the role of the one number that predicts growth, reported to a board, acted on by no one.

I'll put my own bias on the table, since I sell software that reads customer comments: I'm the founder of FeedSense, an AI feedback-analysis tool. That's precisely why I won't pretend a composite number is doing work it isn't. The score was never the point. The three sentences a customer typed underneath it were always the point, and most of the industry built dashboards that foreground the number and bury the sentences. You can fix that with a spreadsheet and a Friday afternoon before you ever pay a vendor.

The position to defend at the leadership table is narrow: keep the NPS survey, stop treating the quarterly company-wide aggregate as research, and re-point attention at the free-text, the segments, the behavioral pairing, and the churner interviews. It's not a budget fight — nobody has to defund anything. It's an attention fight, moving the board's gaze from the number that's easy to read to the comments that are hard to read and worth reading. If your team runs an NPS program and the score has never once changed the roadmap, you don't have a measurement problem. You have a ritual that produces a slide. The data was never the theater. The slide was.

Frequently Asked Questions

Is NPS worth it for SaaS companies?

NPS is worth it as a diagnostic input, not a board-deck headline. The score alone, reported company-wide each quarter, rarely changes a roadmap. The value is in the open-text comments, journey-stage segmentation, and pairing the score with behavioral data. Used that way it's a genuine research instrument. Used as a single aggregate number, it's mostly decoration.

What are the main problems with the NPS score?

NPS has four structural problems as a ritual metric: sample bias (only ~13% of B2B customers respond, skewing toward extremes), cross-cultural skew (a 9 means different things in the US versus Japan), the score-vs-action gap (the number tells you nothing about what to fix), and category manipulation (teams hit targets by surveying only happy cohorts). Each undermines the company-wide aggregate independently.

Does NPS actually predict company growth?

The evidence is weak. The landmark 2007 Journal of Marketing study by Keiningham and colleagues found "no support" for the claim that NPS is the single most reliable indicator of a company's ability to grow, and failed to replicate its "clear superiority" over other metrics. A balanced MeasuringU assessment graded the growth-prediction claim a C+ — useful, but nowhere near "the one number you need to grow."

Why is NPS different across countries?

Because the 0-10 scale is read through cultural norms about extreme ratings. SurveyMonkey's global data shows Brazil at 62 and Japan at -52 for comparable products — a 114-point gap driven largely by response style, not satisfaction. Peer-reviewed research finds East Asian respondents avoid scale extremes far more than Americans, so a global NPS average mixes incompatible scales and produces misleading cross-market comparisons.

What are good alternative metrics to NPS?

The best alternatives aren't single metrics but a practice: read and theme the NPS free-text comments, segment scores by journey stage, pair sentiment with a behavioral metric (retention at N days, feature adoption, usage frequency), and run qualitative churner interviews. Behavioral metrics are harder to game than surveys. For a full breakdown of the survey-metric landscape, see the NPS vs CSAT vs CES comparison.

How is NPS gamed or manipulated?

Teams inflate NPS by controlling who gets surveyed and when. Tactics documented by CX firm Satrix Solutions include surveying only satisfied cohorts, timing surveys after positive interactions, excluding likely churners, and tying bonuses to the score. Even NPS creator Fred Reichheld conceded in 2021 that linking scores to compensation invites "pleading, bribery and manipulation." Because the number depends on the sample, the cheapest way to raise it is to change who you ask.

Should I stop running NPS surveys entirely?

No. Keep the survey, change what you do with it. The data is useful; the quarterly company-wide ritual is the problem. Stop reporting the single aggregate as research, and re-point attention at the open-text comments, journey-stage segments, behavioral pairing, and churner interviews. It's an attention shift, not a budget cut. The metric earns a modest place as a trend line — just not the role of the one number that predicts growth.

What's the single most useful thing to do with NPS data?

Read the free-text comments and report the themes, with the score as a footnote. The number tells you sentiment moved; the comment tells you what to fix. Product expert Itamar Gilad calls the "What's the main reason for your score?" follow-up the most valuable part of the survey. A board that hears a specific, themed problem can decide; a board that hears "NPS is up two points" cannot.