The customers who complain are not your biggest problem. The ones who say nothing are. Here is how to find them in data you already hold.
Every customer experience programme is built around the people who speak. Complaints route to a queue. Detractor scores trigger an alert. A furious verbatim gets read out in the Monday meeting. All of that is useful, and almost none of it is where your churn is coming from.
The customers who complain are talking to you. That is a relationship, not a risk. You can fix things for people who tell you what is wrong, and a good proportion of them stay because you did.
Your problem is the middle of the distribution, and the people who leave without ever appearing on a list.
The 40% your dashboard calls fine
Passives, the customers who score you 7 or 8, average around 40% of a customer base across industries (Retently and CustomerGauge, 2024/2025 NPS Industry Benchmarks). They are the largest single group in most bases, and they are the group the reporting is designed to skip.
They trigger no alerts, because they are not negative. They appear on no escalation list, because they have not escalated. They are not counted in your headline number at all: the NPS calculation subtracts detractors from promoters and treats the whole passive segment as though it were not there.
Then they behave like this. Repurchase rates for passive customers run 50% lower than for promoters (Bain, The Economics of Loyalty).
Forty per cent of the base, half the repurchase rate, and no line in the reporting. That is not a rounding error at the edge of the data. If 40% of your customers are invisible to the system that sets your priorities, every priority you set is drawn from roughly 60% of the picture.
The most uncomfortable number in customer experience
It belongs to Fred Reichheld, who created the Net Promoter Score in the first place.
Between 60% and 80% of customers who defected had described themselves as satisfied or very satisfied on their last survey (Bain, The Ultimate Question 2.0).
That is worth reading twice, because the instinct is to hear it as a data-quality problem. It is not. These are not detractors who slipped through the net, and they are not people who lied on a form. They were satisfied. The survey recorded them correctly. They left anyway.
Satisfaction measured at a point in time, on a scale, is simply not the same variable as intention to stay. The two track each other loosely enough that a programme can run for years on the assumption that they are the same thing, and then one day you check.
Silence is the default, not the exception
Only 3 in 10 customers will directly tell a brand when something has gone wrong (Qualtrics, 2026 Consumer Experience Trends). The rest just leave.
The silence is also deepening. 30% of consumers now stay silent after a bad experience, an all-time high since 2021, in the same research.
The convenient explanation is feedback fatigue: people are asked too often and have stopped bothering. The evidence does not support it. 81% of consumers say they provide feedback when asked, and only 14% say they provide feedback only after negative experiences (Gartner, How to Boost VoC Survey Response Rates, G00797253). Customers are willing, they are willing more often than most teams assume, and they are willing when things go well rather than only when things go badly.
So the problem is not that customers will not talk. It is that the moments and mechanisms you use to ask are not catching them, and that a customer who has already decided to leave has less reason than anyone to complete one more form.
The gap that lets a leadership team miss all of it
80% of chief executives believe their company delivers superior customer service. 8% of their customers agree (Bain, The Delivery Gap).
That 72-point gap is how quiet churn survives inside otherwise competent organisations. It is not that nobody is measuring. It is that the measurement is read by people whose starting assumption is already 80%, and a number that says “mostly fine” confirms it. Silent leavers live inside the confirmation.
Something you can do this week, with data you already have
Stop looking at the customers who complained. Look at the ones who left.
Step one. Pull every customer who churned in the last twelve months. Cancellations, non-renewals, lapsed accounts, whatever churn means in your business.
Step two. For each one, find their last survey response before they went. The score, the date, and the verbatim if there is one.
Step three. Sort them into four piles. Scored you 9 or 10. Scored you 7 or 8. Scored you 6 or below. Never responded at all.
Step four. Count the piles.
Most teams who run this find that the 6-or-below pile, the one the entire alerting system is built around, is the smallest of the four. If the majority of your churned customers scored you 7 or above, or never answered, then your early warning is calibrated to a signal that fires after the decision has already been made.
It is a half-hour query for anyone with access to both systems. It is also, in our experience, the most persuasive slide a CX team can build in a quarter, because it does not argue from benchmark data about other companies. It uses your own customers to show that satisfaction and safety are not the same thing.
If you want a fifth step, read the verbatims attached to the 7s and 8s who left. Not the scores. The words. They are rarely blank, and they rarely say “satisfied”.
What was in the words that was not in the score
That is usually where the answer to the exercise sits.
A score compresses an experience into one digit. The comment beside it does not. A customer who writes “it was fine, took a while, the third time I called someone finally sorted it” has told you about resignation, low-grade frustration and eroding confidence, and then given you a 7 because the problem did eventually get fixed.
Three broad buckets cannot hold that. Positive, negative and neutral put a contented customer and a customer who has quietly given up on you improving into the same pile, because both are “not negative”. The difference between those two people is the entire difference between a renewal and a departure.
Emotion is the layer underneath. Adoreboard reads the emotion in open-text customer feedback across 24 distinct signals, eight emotions measured at three intensities each, which is what makes disappointment separable from anger and resignation separable from calm. Those signals compound into a Trust Score, and Trust moves before the behaviour does.
In a validation study of more than 20,000 customer records from a major North American telecommunications provider, emotion-based measurement predicted churn with 86% accuracy, landing within 1% of actual historical churn. The prediction was available before the cancellations, not after them.
That is what turning customer experience into revenue impact looks like at the quiet end of the distribution. Read everything, including the neutral responses most tools skip. Rank the issues by what they are costing rather than by how often they are mentioned. Get the warning while there is still a quarter left to act in.
Your loudest customers are already telling you what is wrong. The expensive ones are not.
Curious what this looks like on your own data? Book a 15-minute walkthrough: adoreboard.com/demo
