Customer experience in the US: 4 years down, then a rise

Customer experience quality fell for four years in a row in the United States and Canada and rose again in 2026, according to Forrester’s Customer Experience Index (CX Index). In 2025, the firm surveyed more than 275,000 consumers about 469 brands in 13 countries and recorded the fourth consecutive decline in North America, with the index at a new all-time low (Forrester, 2025). The 2026 edition brought the first year-over-year increase since 2021: among 225 US brands evaluated in both years, 26% improved significantly and 7% declined, with scores still slightly below 2024 levels (Forrester, 2026b). For anyone leading a B2B operation, this consumer index works as a thermometer, and the decision depends on a different question: which experience measure predicts renewal and expansion in your customer base, and where it is recorded.

When a leadership team asks me whether customer experience is worth the investment, it usually has a quarterly NPS on the table, a satisfaction survey after each ticket and no cross-reference between those scores and who renewed or cancelled. The score goes up or down, the presentation celebrates or laments, and no account decision changes. I gathered here the studies with published sample and method on the link between experience and results, the points where they disagree and a roadmap to measure, record and act.

What does the Forrester CX Index show about customer experience?

The CX Index measures how well a brand’s experience strengthens customer loyalty, across three dimensions the firm calls effectiveness, ease and emotion. Forrester asks consumers about brands in industries such as banking, insurance, airlines, retail and government and compares the same brand from one year to the next. The index is useful for reading trends, and the share of brands that improve or decline significantly matters more than the score of a single brand.

Four years of decline in North America

In 2025, in the United States, 25% of the brands evaluated in 2024 and 2025 had significant losses and only 7% improved; in Canada, 18% fell and 1% rose (Forrester, 2025). One quarter of the North American brands that fell in 2025 had also fallen in 2024. The lead analyst attributed the decline to weaker employee experience, less focus on the customer, disappointing technology implementations, AI included, and the gap between what executives think of their own experience and what customers perceive.

That last factor shows up in many B2B operations I assess. The company reports a high NPS and loses large accounts without warning, because the person who answered the survey was the satisfied user and the person who decided to cancel was the director nobody ever asked.

The first rise since 2021

The 2026 edition analyzed more than 224,000 consumers, 462 brands, 13 industries and 13 countries (Forrester, 2026a). In North America, 26% of brands had significant gains, 7% declined and 68% stayed flat, with declines in airlines, credit card issuers and the US federal government. In Europe, 91% of 72 brands stayed flat, and in Asia Pacific 19% of 57 brands fell. In the United States, all three dimensions improved, and emotion had the largest gain, 1.4 percentage points (Forrester, 2026b).

The analyst himself wrote that the widespread rise makes it hard to pinpoint which initiatives worked, and Forrester says it will take several years of progress to confirm the turnaround. For your planning, the useful data sits in the years of decline: from 2022 to 2025 companies invested heavily in service technology, and the firm listed disappointing implementations among the causes.

Does NPS predict growth? Reichheld’s study and its critics

The Net Promoter Score was born from a Frederick Reichheld article in Harvard Business Review in 2003 and became the experience metric most used in boardrooms. The evidence behind it deserves a reading before it becomes a bonus target.

What Reichheld measured in 2003

Reichheld and colleagues gave a battery of questions to customers in six industries, such as financial services, telecom and auto insurance, and matched the answers to purchase and referral histories. With data from more than 4,000 customers, they built 14 case studies, and the question “how likely is it that you would recommend the company to a friend or colleague?” was the best predictor of repurchase and referral in most industries (Reichheld, 2003). In a second stage, Satmetrix tracked customer scores for more than 400 companies and compared them with revenue growth, and in airlines the article found a strong correlation between NPS and average growth from 1999 to 2002.

The article lists its own exceptions: recommendation did not predict growth in industries dominated by monopolies, and some niche companies grew faster than their NPS would suggest. Two method notes matter for anyone using the text as an argument. Satmetrix, which supplied part of the data, was a maker of feedback management software, and the growth analysis is a correlation across companies, without controls for other causes.

What the academic review found

Timothy Keiningham and coauthors tried to replicate Reichheld’s analyses in the Journal of Marketing with longitudinal data from 21 firms and more than 15,500 interviews from the Norwegian Customer Satisfaction Barometer, compared with the American Customer Satisfaction Index, in the industries Reichheld cited as examples. They failed to replicate the “clear superiority” of NPS over other loyalty measures (Keiningham, Cooil, Andreassen and Aksoy, 2007).

A year earlier, Neil Morgan and Lopo Rego had tested six satisfaction and loyalty metrics against the financial performance of US firms between 1994 and 2000. Average satisfaction had the greatest value in predicting future performance, and metrics based on recommendation intentions, NPS among them, had little or no predictive value (Morgan and Rego, 2006).

I read this debate with less drama than the academic dispute suggests. The recommendation question remains useful as a relationship thermometer and loses value when the company treats it as the single number that explains growth. The studies disagree on which metric predicts best, and none of them excuses leadership from testing the link between score and behavior in its own customer base.

Delight the customer or reduce effort?

In 2010, Matthew Dixon, Karen Freeman and Nicholas Toman, of the Corporate Executive Board, published in Harvard Business Review a study of more than 75,000 B2C and B2B customers about recent interactions by phone, voice prompts, web, chat and email, in North America, Europe, South Africa, Australia and New Zealand (Dixon, Freeman and Toman, 2010). The authors controlled for variables such as type of issue, company size, switching costs, price and perceived product quality. The conclusion contradicted the strategy of 89 of the 100 service leaders surveyed: exceeding expectations makes customers only marginally more loyal than solving their problem, and reducing the effort they make carries more weight.

The study also shows why satisfaction misleads. Eighty percent of service organizations used customer satisfaction (CSAT) as their primary metric, and 20% of “satisfied” customers said they intended to leave the company. The authors compared CSAT, NPS and a new metric, the Customer Effort Score (CES), which asks how much effort the customer put in to handle the request, on a scale from 1 to 5. Satisfaction was a poor predictor, NPS did better, and CES outperformed both in service interactions.

Customer group in the study	Indicator	Result
Low-effort interaction	Intention to repurchase	94%
Low-effort interaction	Intention to spend more	88%
Low-effort interaction	Intention to speak negatively about the company	1%
Hard time solving the problem	Intention to speak negatively about the company	81%
“Satisfied” customers	Intention to leave the company	20%
“Dissatisfied” customers	Intention to stay	28%

Source: Dixon, Freeman and Toman, 2010. All figures are intentions stated by the customer.

The authors note that NPS works well as a gauge at the company level and that CES performs better at the interaction level. The distinction resolves much of the apparent contradiction with Morgan and Rego, who measured whole firms over several years: effort explains the ticket, and recommendation and average satisfaction describe the relationship with the brand.

A widely repeated figure says that 96% of customers with a high-effort interaction become more disloyal, against 9% of those with low effort. It appears on blogs and vendor pages attributed to Gartner, which absorbed the Corporate Executive Board. The 2010 article does not contain that number, and I could not open the original document with sample and method, so it stays out of this text.

In B2B, does customer experience lag behind?

A 2016 McKinsey article by Nicolas Maechler, Sanjeev Sahni and Martine van Oostrum states that B2C companies typically score in the 65% to 85% range on customer experience indexes, while B2B companies average less than 50% (McKinsey, 2016). The text reports that broad experience transformations brought reductions of 10% to 20% in cost to serve and revenue growth of 10% to 15%. It does not say which index it used or how many companies entered the comparison, and I treat these figures as a consulting account.

The most useful observation for B2B sits outside the percentages. A business customer’s experience runs through the buyer who negotiates, the user who operates and the director who approves the renewal, and a survey that only hears the main contact measures one role and leaves out the person who decides. Dixon’s study cites a B2B manufacturer that lowered its CES from 2.8 to 2.2 by changing the language its reps used, a sign that the effort metric also works for business-to-business service.

Summary table of customer experience research

Source	Year	Sample and method	Main finding
Forrester, CX Index	2025	More than 275,000 consumers, 469 brands, 13 countries	Fourth straight decline in North America; in the US, 25% of brands fell and 7% improved
Forrester, CX Index	2026	More than 224,000 consumers, 462 brands, 13 countries	First rise since 2021 in the US; 26% of brands improved and 7% fell
Reichheld, Harvard Business Review	2003	More than 4,000 customers in 14 cases across 6 industries; scores from more than 400 companies	Recommendation predicted repurchase and referral; NPS correlated with growth in several industries
Morgan and Rego, Marketing Science	2006	US satisfaction index and financial performance, 1994 to 2000; six metrics	Average satisfaction predicts best; recommendation had little or no predictive value
Keiningham and coauthors, Journal of Marketing	2007	21 firms and more than 15,500 interviews in Norway, longitudinal data	Did not replicate NPS superiority
Dixon, Freeman and Toman, Harvard Business Review	2010	More than 75,000 B2C and B2B customers in remote service channels	Low effort linked to 94% repurchase intention; CES beat CSAT and NPS in service
McKinsey	2016	Consulting article, no sample disclosed	B2B below 50% on experience indexes, against 65% to 85% for B2C

What these data do not say

The CX Index measures consumers in North America, Europe and three Asia Pacific countries, without Brazil, in consumer and public service industries. None of them resembles selling software, recurring services or industrial inputs to companies, and the 2026 rise is no forecast for your accounts. The full methodology sits in paid reports, and the figures in this text come from the analysts’ open publications.

Dixon’s study measures stated intention, and the same person rated the effort and their own loyalty, which may inflate the relationship between the answers. The authors worked for a firm that sold research and advisory services to service organizations, and the study is sixteen years old, predating business messaging apps and AI agents.

In the NPS debate, Reichheld and his critics studied whole companies and consumer industries, and none of the three papers tests the recommendation question in B2B accounts with several decision makers. The link between score and growth is a correlation in all three: a growing company may have more satisfied customers because it invests in service with the cash that growth produces. The McKinsey article is a consulting piece with no published method, and its revenue gains come from cases the authors chose.

How to measure B2B customer experience and record it in the CRM?

The research points to one design principle: each metric at the level where it predicts best, with the answer stored where the decision happens. A score that lives in a separate survey tool, with no link to the contact and the company in the CRM, never reaches the account review, and CS finds the detractor on the day of the cancellation.

Metric	Level	When to ask	Who answers	Where to record in the CRM	Decision it triggers
Effort (CES)	Interaction	When the ticket closes and at the end of onboarding	Whoever opened the ticket or took part in the implementation	On the ticket, linked to contact and company	Review of the process that caused high effort
Satisfaction (CSAT)	Delivery	After project milestones	User and delivery owner at the client	On the project, with date	Correction before the next milestone
Recommendation (NPS)	Account relationship	Every six months and 90 days before renewal	Decision maker, main user and buyer	On the contact record, with role, and rolled up to the company	CS leadership calls detractors within a set deadline
Behavior	Account	Monthly, no question asked	Nobody; it comes from usage and finance	On the company, with history	Test of which scores predict renewal

The last row is the one almost nobody does and the one that settles the academic debate for your company. With twelve months of scores by account and by role, you compare renewal and expansion among promoters, passives and detractors and find out whether NPS predicts anything in your base or whether service effort explains cancellations better. The risk signals that feed this analysis are in B2B churn, and the account design that receives these scores is in B2B post-sale.

The cross-reference depends on a base where contact, company, role and renewal date are filled in and free of duplicates, the work I describe in CRM data quality. If your service team already has AI agents, measure effort in the conversations they handle too, as I discuss in AI in customer service.

What leadership does with these numbers

The studies do not deliver a winning metric, and a leadership team waiting for that verdict ends up choosing by the preference of whoever presents. They deliver a method, and the roadmap I use with CS, sales and service leaders has six steps.

1. List what you already measure. Gather the surveys sent today, with question, timing, audience and tool, and note which answers reach the CRM linked to the contact and the company.
2. Separate interaction from relationship. Use effort at the end of tickets and onboarding and recommendation or satisfaction for the account relationship, without asking one score to answer both questions.
3. Hear more than one role per account. Define which roles answer the relationship survey, with at least the decision maker and the main user, and roll up the company score without erasing each person’s score.
4. Close the loop with a deadline. Write down who calls the detractor and the high-effort customer, within how many business days, and record the conversation and the action in the CRM.
5. Test the metric in your base. After twelve months, compare renewal and expansion by score group and keep on the leadership dashboard the metric that showed a link to behavior.
6. Take the score out of the bonus until you have the test. While the company does not know whether the metric predicts revenue, targets tied to it push the team to choose who gets the survey.

Customer experience and L&D: what HR needs to assess

The request usually reaches HR as customer delight training, and Dixon’s research suggests a different scope. The team that moves loyalty solves the problem with less effort for the customer, anticipates the next question and records what happened, and those behaviors are taught with real cases from the operation. To measure the effect, compare before and after the CES of tickets, the share of detractors contacted on time and the share of answers linked to contact and company in the CRM.

The data show that experience connects to revenue when the company measures each level with the right metric, records the answer where the decision happens and tests that link in its own base. That is the work of the in-company CRM adoption training, run with leaders and CS, service and sales teams in the CRM the company already uses. The cohorts finish with the CRM best practices guide, which defines where each score and each signal is recorded, and the portfolio and database management manual, which states who acts on each answer and within what deadline.

Frequently asked questions

Has customer experience gotten worse in recent years?

In North America, yes, for four years in a row through 2025, according to Forrester’s CX Index. In 2026 the US index had its first rise since 2021, with 26% of brands improving and 7% declining, and scores still slightly below 2024. The index measures consumers and does not include Brazil.

Does NPS predict company growth?

Reichheld’s 2003 article found a correlation between NPS and growth in several industries, with exceptions such as markets dominated by monopolies. Academic studies from 2006 and 2007 did not replicate NPS superiority over other metrics, and one of them found little or no predictive value. The safe way to know is to test the link between score and renewal in your own base.

What is the difference between NPS, CSAT and CES?

NPS asks how likely the customer is to recommend the company and measures the relationship. CSAT asks how satisfied they were with a delivery or interaction. CES asks how much effort they put in to handle the request, and in the study by Dixon, Freeman and Toman it predicted loyalty best in service interactions.

How do you measure customer experience in B2B?

Measure effort at the end of tickets and onboarding, measure the account relationship every six months and before renewal, and hear at least the decision maker and the main user. Record each answer in the CRM, linked to the contact, the role and the company, so you can cross it with renewal and expansion.

Is the figure that 96% of high-effort customers become disloyal reliable?

It circulates attributed to Gartner, which absorbed the Corporate Executive Board, author of the 2010 study. The article published in HBR does not contain that number, and I could not find the original document openly available with sample and method. Until I see the source, I prefer the figures from the 2010 article, which are intentions stated by more than 75,000 customers.

How do you prepare the team to improve customer experience?

Start with the tickets and stages that cause the most effort, using real cases from the operation, and train the team to solve the problem and anticipate the next question. Include CRM recording and deadline-bound follow-up with detractors in the training, because without that record leadership cannot measure whether behavior changed.