NPS Was Never Meant for This

I was in a shop recently. Small posters at every customer service desk. QR codes. “Rate us!” On a scale of 0 to 10, how likely are you to recommend us?

I asked the woman at the counter: are your bonuses connected to this score?

No, she said. We have personal sales targets. But our manager does.

So the person closest to the customer has no stake in the number. The person furthest from the customer has their bonus tied to it. And the customer is being asked to perform unpaid labor for a metric that flows upward through the org chart without touching anyone who could act on it.

This is Net Promoter Score in 2026. The man who invented it would not recognize what it became.

One number to rule them all

“How likely is it that you would recommend this company to a friend?” Fred Reichheld proposed this single question in 2003. His claim: it correlated with revenue growth better than any satisfaction survey. Willingness to recommend carries personal reputational risk, which makes it a stronger signal of genuine loyalty than “are you satisfied?”

The insight was real. The metric was a proxy. One question instead of a complex satisfaction survey.

What Reichheld did not propose: tying the score to employee bonuses, comparing teams by it, reporting it on earnings calls, or building performance reviews around it. The metric was designed to listen. Organizations turned it into a target.

The ratchet

Proxy metrics are corrupt in a predictable way. Goodhart named it in 1975: when a measure becomes a target, it ceases to be a good measure. But the corruption itself is the boring part. The interesting question is why it persists.

NPS followed the path cleanly.

First, the proxy was genuine. “Would you recommend?” really did correlate with customer behavior. At this stage, NPS was useful. Organizations that tracked it learned something about their customers.

Then the score acquired consequences. Companies tied NPS to executive bonuses, team KPIs, and frontline evaluations. The moment the number carried career implications, the rational response shifted. “Make the customer happy” and “get the customer to rate us high” sound like the same thing. They are not the same thing. They diverge the moment someone’s paycheck depends on the difference.

Gaming followed, as it always does. Companies timed surveys to reach only satisfied customers, pre-framed responses (“anything less than a 9 counts as a failure for our team”), and excluded segments likely to score low.

At car dealerships, it became score-begging. Fortune reported salespeople telling customers: “If you don’t give me a 10, I won’t be able to feed my kids.” This was the rule, not the exception, in U.S. auto retailing.

One company improved its NPS by double digits without changing anything about its product. They placed a customer support phone number above the survey question. Detractors called instead of completing the survey. The score went up. Nothing else changed.

Gaming targets are a rational response to irrational system design, as John Seddon put it. If you reward hitting a number, people will hit the number. Whether that serves the customer is a question the system never asks.

And then came lock-in. NPS embedded itself in board reporting, quarterly earnings calls, vendor contracts, compensation models, and entire SaaS platforms. Two-thirds of the Fortune 1000 adopted it. SAP acquired Qualtrics for $8 billion. Medallia sold for $6.4 billion. The NPS software market alone is $1.4 billion. Careers were built on “I improved NPS by 20 points.” Proposing to retire the metric now threatens the infrastructure and the people built on it.

The metric cannot go backward. Each stage created new stakeholders who benefited from the increase in the number, regardless of whether customer experience actually improved. That’s the mechanism.

The metric was pointing the wrong way

Research from Bain & Company, the firm where Reichheld developed NPS, shows the causal direction he got backward. Companies that deliver well on what customers care about, quality, simplicity, time savings, risk reduction, produce more than double the NPS and four times the revenue growth of companies that don’t. Value drives NPS. NPS does not drive growth. The metric was never a leading indicator. It was a lagging echo of something organizations already controlled: whether they delivered value.

The academic consensus caught up over the course of two decades. Keiningham replicated Reichheld’s methodology in 2007 and found plain customer satisfaction predicted growth just as well. Dawes confirmed it in a 2024 review. Reichheld’s foundational study was never peer-reviewed. The underlying data was never published.

And NPS doesn’t even measure what it claims to measure. It measures what people say they will recommend, not what they do. A customer who scores 9 may never refer anyone. In markets with high switching costs, people report satisfaction not because the product is good, but because leaving is painful. NPS conflates genuine satisfaction with captive tolerance.

The inventor walks away

In 2021, Reichheld published “Net Promoter 3.0” in Harvard Business Review. He cataloged the corruption of his own metric: “pleading (‘I’ll lose my job if you don’t rate me a 10’), bribery (‘we’ll give you free oil changes for a 10’), and manipulation (‘we never send surveys to customers whose claim was denied’).” Companies publicly report unaudited, self-reported scores. No safeguards, no methodology explanation, no verification.

NPS is “used incorrectly 90% of the time,” he wrote. His longtime collaborator Rob Markey drew the line: “The score is almost worthless, and often even harmful, if not embedded in a system of listening deeply to customers.”

And then Reichheld introduced a replacement. Earned Growth Rate tracks actual repeat purchases and referred customers, based on audited revenues, not survey responses. It measures exactly what NPS was supposed to proxy for. He built it because his original invention was too easily gamed.

The man who created “the one number you need to grow” spent eighteen years watching organizations turn it into a number that measures nothing but their ability to manipulate a survey.

What the poster at the service desk actually measures

Back to the shop. The poster with the QR code. The employee who doesn’t benefit from the score. The manager whose bonus depends on it.

That poster doesn’t measure loyalty, doesn’t predict growth, and doesn't tell anyone whether value is being delivered. It measures the organization’s ability to get customers to scan a code and enter a number that flows into a dashboard, where a manager needs it to be green so their quarterly review goes well.

NPS is not special. Every proxy metric that acquires consequences follows the same ratchet. Velocity. Story points. Code coverage. Employee engagement scores.

The number your organization treats as real deserves the same question I asked at the counter: whose bonus depends on it, and are they anywhere near the work?