NPS Isn't Your Revenue Problem. The Missing Alignment Is.

Customer Experience ExcellenceCustomer Experience ManagementCustomer Value PropositionCustomer Based Selling

Nov 6

In my last piece I argued that brand measurement is not broken, the brief is. Brand is the promise. Customer experience is the delivery. Both get judged on numbers that finance does not trust.

CX has the same problem, one function over. Teams report NPS and CSAT. Finance counts revenue. Nobody agreed upfront what evidence would connect the two.

McKinsey’s survey of CX leaders put numbers on it: 93% rely on a survey metric as their primary measure of CX, yet only 4% say they can calculate the ROI of CX decisions.

The disconnect is real

Only 15% of those leaders were fully satisfied with their measurement, and the typical survey reached about 7% of customers. Gartner’s research points the same way: only about half of firms had financial metrics in place to measure CX ROI, rather than informal or nonfinancial ones.

One executive at a large financial services company told McKinsey that many parts of the organisation simply claim a business impact from their CX initiatives “with no real evidence.” That is the CX version of “awareness” as a brand outcome: a metric standing in for a result.

Why scores do not equal revenue

The academic evidence is more useful, and less flattering to NPS, than most CX decks suggest.

NPS was oversold. Reichheld’s 2003 claim that it is the single best predictor of growth has not held up. A 2007 replication across 21 firms and more than 15,500 interviews found no support for it, and Reichheld himself conceded in 2021 that unaudited, self-reported scores had undermined NPS.
Satisfaction was undersold. In the best customer-level study of observed retention, top-box satisfaction was the strongest predictor, official NPS came next, and the gap between them was not substantial. No single metric won across industries.
Intentions are not behaviour. The popular finding that 94% of low-effort customers intend to repurchase measures what people say they will do. In the same retention study, customer effort was the one metric with no significant effect on what customers actually did.
Rank beats score. Absolute satisfaction explains little of a customer’s share of wallet. How they rank you against the other providers they use explains much more. In banking and insurance, where customers hold products with several providers, that matters. (The authors commercialised the method, so treat the correlations as strong but not fully independent.)

Three traps sit underneath all of this. Reverse causation: healthy firms can afford better service. The say-do gap: stated intent overstates behaviour. Sample bias: the customers about to leave are the least likely to answer the survey.

Is it realistic? Yes, with conditions

It is realistic to tie CX to revenue credibly, but not by proving that NPS drives revenue. What you can prove is that specific experiences, measured at customer level, change behaviours finance already values.

What you can credibly show:

Retention and churn differences between customers who had a good versus a poor experience on a specific journey
Revenue at risk from failed journeys
Cost-to-serve changes, such as repeat contacts and rework
Incremental revenue from a CX intervention, with a control group

What you cannot: that a one-point NPS rise caused a specific revenue gain without a comparison group, or any borrowed “one point equals $X million” model applied to your business without calibration.

It works best where you can see customers leave: banking, insurance, health insurance, telecoms and subscriptions. Those are the sectors best placed to get this right.

Who has done it

The pattern in credible cases: customer-level data, a comparison group, and an outcome finance already tracks.

T-Mobile US tested its Team of Experts model for more than two years before national launch, then reported churn and cost-to-serve alongside NPS. The company says postpaid phone churn fell 39% and calls per account fell 37% since rollout began. Pricing and network moves ran in parallel, so the CX share is not isolated.
DBS had the lowest satisfaction scores among Singapore banks in 2015, then reorganised around customer journeys. It reports about S$1 billion in economic value from data and AI, and secondary reports say outcomes are compared against control groups. Customer outcomes also sit in its executive scorecard. Figures are company-reported.
Telstra shows the limits of putting NPS in executive pay. It had internal audit verify episode NPS, and relationship NPS was later removed from executive incentives, according to remuneration analysts. Episode scores respond to operational fixes. Relationship scores are much harder to control.
Medicare Advantage is the clearest case because the revenue formula is explicit. Plans rated 4 stars or above earn a quality bonus payment, and BCG estimated that moving from 3.5 to 4 stars is worth roughly US$400 per member per year. When finance can see the formula, CX becomes a revenue line.

A note on evidence quality: most company figures are self-reported and unaudited, and I could not find a primary disclosure from a major APAC insurer linking CX to financial results.

How to do it: build one customer-level join

Pick the journey with the most revenue at stake. In insurance that is usually claims or renewal. In banking, onboarding or problem resolution. Then link experience data to retention and value at customer level. Here is how it works for insurance claims.

1. Fix the question. Agree one sentence with finance: do customers with a poor claims experience renew at a lower rate than comparable customers with a good one, and what is that worth?

2. Build the data spine, one row per customer.

Data spine for a claims and renewal analysis
Layer	Examples	Typical source
Customer and policy	Tenure, age band, product, premium, channel, region, prior claims	Policy admin, CRM
Experience	Days to settle, number of contacts, rework, complaint flag, claim outcome, survey score where available	Claims system, contact centre, survey platform
Outcome	Renewed or lapsed, premium after renewal, products added	Billing, policy admin

Define experience from operational signals first, such as slow settlement or three or more contacts. Surveys reach few customers, and the people most likely to leave are the least likely to answer. Use survey scores to check that your definition matches how customers felt.

3. Set the groups before you look at results. Agree good and poor thresholds with the claims team up front, so nobody can cherry-pick.

4. Compare like with like. Control for what else drives churn: claim size, tenure, product and, above all, the renewal price increase. Report the adjusted gap, not the raw one.

5. Convert to dollars. Adjusted churn gap × customers in the poor group × average annual premium. Finance usually prefers margin to premium, so show both. Here is an illustration with made-up numbers: 10,000 customers had a poor claims journey, and their adjusted renewal rate is 7 points lower. That is 700 extra lapses × S$2,000 average premium = S$1.4 million in annual premium at risk. Add the cost-to-serve difference as a second line, because it sits directly in the cost base.

6. Be honest about what it means. The number is revenue associated with poor experience, which is an upper bound on what a fix could recover. To prove cause and effect, fix the slowest part of the claims journey for one region or a random group and compare against a control. That is the DBS and T-Mobile pattern.

Practicalities. Get privacy and compliance sign-off early (PDPA in Singapore, plus any sector rules). One analyst, one data engineer and one finance partner is enough, and 8 to 12 weeks is realistic for a first version.

What finance will and will not accept

Finance leaders are not hostile to CX. They are hostile to claims they cannot test.

What finance usually accepts and rejects as proof of CX value
Usually accepted	Usually rejected
Retention and revenue differences on your own customer data	Cross-industry correlations
Pilot versus control results	Vendor ROI calculators and “one point equals $X million” models
Cost-to-serve savings tied to the cost base	Projections built on stated intentions
Data audited by internal audit	An NPS trend placed beside a revenue trend

Two shared outcome metrics help. The first is earned growth, an accounting-based measure that combines net revenue retention with revenue from customers won through referral. Reichheld introduced it precisely because self-reported NPS had become easy to game. The second is relative rank, which asks customers which providers they use and how they rank you against them. In multi-provider categories, rank tells you more about share of wallet than any absolute score.

Back to brand: one case for the CFO

This is where the brand and CX arguments meet. Brand sets the expectation. CX delivers or breaks it. The gap shows up in behaviour finance already tracks: retention, share of wallet, price realisation, referral and cost-to-serve.

Forrester’s analysis of US direct banks suggests the two reinforce each other. Improving both brand experience and customer experience was associated with a 2.3x revenue lift, versus 1.5 to 1.6x for improving either alone. That is analyst modelling, not proof, but it points the right way.

So make one case, not two. Marketing, CX and finance agree the outcome, the evidence standard and the comparison group before the money is spent.

The CX scorecard does not need a better survey. It needs a join to the ledger, and a CFO who helped design it.

Start small. Pick one journey, build one join, run one controlled test. Then you have a number finance can challenge, which is the only kind of number they will trust.

Sources

Figures are as reported in the pages below. Company results are self-reported and unaudited, and several sources have a commercial stake in the method they describe (Bain co-owns NPS, Gartner sells the effort score, the share-of-wallet authors commercialised their method).

customer experiencecustomer insightsrevenue alignmentcustomer centricity

Jaslyin Qiyu