The State of Voice of Customer 2026

The business case for listening to customers has always rested on retention economics. The most-cited version: acquiring a new customer costs far more than keeping one — Harvard Business Review puts the range at five to twenty-five times, framed explicitly as a synthesis of studies rather than one controlled experiment.

The companion figure comes from Bain's Frederick Reichheld: increasing retention by five percent can lift profits by 25% to 95%. It is genuinely attributable to Bain — but the underlying research traces back to a 1990 HBR paper, and the range is a cross-industry generalisation, not a single measurement. Cited honestly, it still makes the point: small retention gains compound into large profit gains, which is exactly why understanding why customers leave is worth doing well.

The problem is that the customers who would tell you why they're leaving mostly don't. Decades-old research from TARP — still the most-cited figure on silent churn, though its primary documents date to the late 1970s and 1980s — suggests only about one unhappy customer in twenty-six ever complains. The rest simply leave. Treat it as illustrative lineage, not a fresh measurement, but the direction is not seriously disputed.

And the cost of the experiences that drive them away is well documented in more recent work. PwC's 2018 survey of roughly 15,000 consumers found that 32% would walk away from a brand they love after a single bad experience, and 59% after several. If most of that dissatisfaction never reaches a survey, a survey-only programme is, by construction, blind to the majority of it.

Surveys are not wrong; they are increasingly insufficient. The clearest long-run evidence of survey fatigue comes from Pew Research Center: response rates to its telephone polls fell from 36% in 1997 to 9% by 2012, and to just 6% by 2018. These are opinion-poll response rates, not CX surveys specifically — but they are the best-documented proxy for a trend every research team now feels.

When only a single-digit percentage of people answer, two problems follow. The sample skews toward the few who bother, and the signal arrives late — a quarterly score can't tell you about a problem that started this week. A programme that leads with solicited surveys is asking a shrinking, self-selecting minority, on a lag.

If customers won't fill in surveys, where is the signal? It is in the text and conversations they generate anyway — support tickets, chat logs, call transcripts, reviews, social posts. A widely-cited estimate — traceable to a 1998 Merrill Lynch figure and echoed since by IDC and Gartner — holds that 80–90% of enterprise data is unstructured. Treat the exact percentage as an industry-consensus estimate rather than a precise measurement, but the implication is real: the overwhelming majority of customer signal is in a format legacy VoC tools were never built to read.

This is the gap modern VoC has to close. The value is no longer in collecting more structured responses; it is in making the unstructured majority countable — extracting themes, sentiment, and intent from language at scale.

What changed is that reading unstructured feedback at scale is now practical. Large language models can classify sentiment, cluster themes, and summarise thousands of comments the way an analyst would — only faster and continuously. The market has noticed: Gartner reported in December 2024 that 85% of customer service leaders planned to explore or pilot a customer-facing conversational generative-AI solution in 2025.

The frontier is moving from analysis to action. The interesting question in 2026 is no longer "can AI read the feedback?" but "can it reason over it and recommend the right next step — with a human still deciding?" That is the shift from dashboards you interpret to systems that propose.

Customer expectations have moved in the same direction. McKinsey's 2021 personalization research found 71% of consumers expect companies to deliver personalized interactions and 76% get frustrated when they don't — and that faster-growing companies drive 40% more of their revenue from personalization than slower-growing peers. Salesforce's State of the Connected Customer has repeatedly found a large majority of customers expect immediate interaction when they reach out.

Personalisation and immediacy both depend on the same thing: knowing what an individual customer needs right now, from their signals, and acting on it. That is VoC's job description in 2026.

Spending is tracking the trend. Grand View Research valued the Voice-of-Customer analytics market at roughly $1.7B in 2024, projecting about $4.7B by 2030 — an ~18.9% CAGR. Market-sizing varies widely by firm and definition (other analysts put the broader VoC tools market far higher), so treat any single figure as one firm's estimate, not a settled number. The direction is consistent across all of them: up, and fast.

For context on how entrenched the incumbent metric is: Fortune, citing Bain, estimates at least two-thirds of the Fortune 1000 use Net Promoter Score. The category is huge and standardised — which is exactly why a shift in how it's done matters.