In 2003, Fred Reichheld had a simple, brilliant idea: measure loyalty with a single question. Twenty years later, NPS is everywhere, cited in every executive committee, written into every BPO contract. And yet customer experience quality has been declining for three consecutive years, according to Forrester. We’ve never measured NPS this much. We’ve never disappointed this much.
In 2003, Fred Reichheld published an article in the Harvard Business Review titled “The One Number You Need to Grow.” The idea was appealing in its simplicity: one question, “would you recommend this company to someone close to you?”, was enough to predict growth better than any battery of satisfaction questions. NPS was born, and it conquered dashboards worldwide.
Twenty years later, something is off. Forrester measured, in 2024, the sharpest decline in customer experience quality in a decade: 39% of American brands recorded a significant CX decline, a record. In 2025, the trend continued: 21% of brands worldwide still declined, only 6% improved. Over this same period, organisations have never tracked their NPS more closely. The paradox is striking: we measure more, we disappoint more. It’s not a measurement problem. It’s a problem of what we measure.
NPS measures a stated intention at a specific moment: the likelihood a customer will recommend your brand to someone in their circle. It's a measure of overall sentiment, not future behaviour. This distinction, seemingly trivial, is fundamental.
A customer can give you an 8 out of 10 on NPS and leave three months later because a competitor offered a cheaper deal, a simpler journey, or a smoother experience. They don't recommend you any less than before. They just left. And your NPS never saw it coming. According to a London School of Economics study, the correlation between NPS movement and actual revenue growth is only 0.24, far from the predictive power Reichheld originally claimed.
Gartner frames the problem precisely: loyalty isn't just tied to how customers feel about a brand, but to how the experience with that brand makes them feel about themselves. Do they feel more competent, more confident, better served? That's what Gartner calls the Value Enhancement Score (VES), a metric NPS doesn't capture. And according to Gartner's data, 82% of customers with a positive VES are retained, and 86% spend more in the future, correlations far stronger than what NPS produces.
In 2003, recommending a brand and staying loyal to it were strongly linked behaviours. The economics of recommendation were simple: if you were satisfied, you said so around you, and people you knew trusted your opinion. Loyalty followed.
In 2026, this mechanism has become more complex along at least three dimensions. First, switching costs have dropped in most sectors. Changing phone provider, bank, insurer, or e-commerce platform takes less time than before. A satisfied customer who recommends you can still leave if they find something more convenient elsewhere. Recommendation and retention no longer predict each other as well.
Next, channel fragmentation has multiplied touchpoints. Loyalty no longer plays out over a single interaction but over the consistency of dozens of interactions across different channels. A customer can have a high NPS based on a positive in-store experience, and leave because their digital experience is poor. The global NPS masks these disparities.
Finally, the behaviour of "passives" (scores 7-8) has become a major blind spot. Passives aren't satisfied: they're indifferent. And indifference, in a competitive market, is a posture of imminent departure. According to data compiled by CMSWire, only 23% of American CX directors still use NPS as their primary KPI in 2025, down from a majority five years ago. It hasn't been abandoned: it's been demoted. Which may be worse.
Silent churn is one of the costliest manifestations of NPS's limits. A satisfied customer, who reports a correct recommendation intention, gradually reduces their usage, decreases their purchases, drifts toward alternatives without ever expressing open dissatisfaction. They don't complain. They leave quietly.
This phenomenon is structurally invisible in NPS because NPS measures stated intention, not observed behaviour. A customer silently churning can still give you a 7 or 8 in your quarterly survey. They would recommend you, if asked. But they're buying less and less. Behavioural data sees it. NPS doesn't.
It's precisely for this reason that Armatis's CX Horizon 2030 study, conducted among CX directors at ENGIE, Volkswagen, SFR, LVMH, MACIF, Matmut, La Banque Postale, and Carrefour, reveals that 85% of decision-makers now see customer service as a profit centre rather than a cost centre. This shift in perspective isn't rhetorical: it reflects an awareness that the real value of the customer relationship is measured in retention, average basket, and CLV, not in recommendation intention.
NPS isn't to be thrown out. It remains useful for benchmarking, long-term trends, and executive communication. What needs correcting is the habit of using it alone, as if a single question could sum up the complexity of a customer's loyalty.
Several complementary metrics are now better correlated with real loyalty.
Customer Effort Score (CES) measures the effort a customer had to make to accomplish their task. It's one of the strongest predictors of churn: according to Gartner data, 96% of customers who experienced a high-effort interaction become more disloyal, versus only 9% for those who had an effortless interaction. Effort predicts departure better than stated dissatisfaction.
Retention rate and Customer Lifetime Value (CLV) are the most direct indicators of behavioural loyalty. They don't measure what the customer says: they measure what they do. In BPO contracts, they're starting to appear alongside CSAT and NPS as outcome commitments, not just tracking indicators.
Gartner's Value Enhancement Score (VES) measures whether the interaction with customer service strengthened the customer's confidence in their purchase and their ability to get value from it. It's an indicator focused on how the customer feels afterward, not during. And its correlation with retention is significantly stronger than NPS's.
The re-contact rate (customers who contact again for the same issue within 48-72 hours) is an often-overlooked operational signal but a highly predictive one for churn. A customer who has to call back twice isn't just frustrated: they doubt. And that doubt will show up in their next NPS score, six weeks later, when it's too late to act.
The problem isn't NPS itself, it's how it's used. When NPS becomes the sole loyalty indicator presented at board level, two things happen: the score gets optimised rather than the experience, and the weak departure signals that accumulate outside what a recommendation survey can capture get missed.
A robust loyalty management system in 2026 rests on three complementary layers. The first is declarative: what the customer says (NPS, CSAT, verbatims). It gives the tone and the trend. The second is behavioural: what the customer does (retention, purchase frequency, CLV, re-contact rate, channel usage). It gives the reality. The third is predictive: signals that precede departure (degraded CES, reduced frequency, smaller average basket, increased self-service contacts). It gives the time to act.
Presented this way at board level, this dashboard says something NPS alone can't say: "Our customers report a stable recommendation intention, but their purchasing behaviour indicates a weakening bond in the 35-50 age segment, particularly on the digital channel." That's actionable. That's what a CX director can defend in front of a CFO or CEO.
The contact centre plays a particular role in this system. It's where the highest-stakes emotional interactions, complaints, and complex requests play out. These are also the interactions that most influence real loyalty, far beyond their weight in total contact volume. A well-managed contact centre, with SLAs on FCR, CES, and CSAT by interaction type, contributes directly to the behavioural layer of loyalty: less re-contact, less effort, less silent churn.
No, but it's being demoted. According to CMSWire 2025 data, only 23% of American CX directors still use it as their primary KPI, down from a majority five years ago. It hasn't been abandoned: it's been integrated into broader dashboards, alongside behavioural and effort indicators. That's a healthy evolution, not a disappearance.
NPS measures stated recommendation intention at a specific moment. It's a measure of overall sentiment toward a brand, not a reliable predictor of future behaviour. It's useful for long-term trends, sector benchmarking, and board communication. It doesn't let you anticipate churn, detect weak signals, or identify friction points in the journey.
No single metric replaces NPS alone: it's the combination that changes things. Customer Effort Score (CES) is the best short-term predictor of churn at the interaction level. Retention rate and CLV measure real behavioural loyalty. The 48-72 hour re-contact rate is a powerful operational signal. Together, these indicators give a far more accurate picture of relationship health than a quarterly NPS.
By structuring reporting into three layers: declarative (NPS, verbatims), behavioural (retention, CLV, purchase frequency), and predictive (CES, re-contact rate, usage-reduction signals). This format distinguishes what customers say, what they do, and what they're about to do. It's more actionable for senior leadership than NPS movement alone.
Yes, but not alone. In serious BPO contract SLAs, NPS is complemented by CSAT by channel, FCR, and re-contact rate. These operational indicators connect contact centre performance to measured loyalty, and detect degradation well before it appears in quarterly NPS. A BPO contract that contracts only NPS as a loyalty measure manages too late.
Sources
Armatis is a European specialist in customer relations and business process outsourcing (BPO), operating across multiple continents with thousands of employees serving companies of all sizes and sectors. The company designs and manages end-to-end customer service operations: multichannel contact centres, complaints handling, technical support, back-office and digitised processes. Backed by integrated technology infrastructure and the ability to adapt to any sectoral and regulatory context, Armatis helps its clients combine operational performance, quality of experience and cost control, wherever they need it.
Contact our teams to discuss your challenges and find out how we can support you
Join the leaders who trust our multilingual and technological expertise.