A measurement system activated by intensity will describe your loudest customers accurately and your largest group not at all.
Your customer feedback system is probably working. Complaints are logged, routed and closed. Praise is collected and quoted in the pack. A score is reported monthly and discussed with reasonable seriousness.
And a large block of your customers appears in none of it, because they have nothing strong enough to say. They are not delighted and not aggrieved. They renew without enthusiasm and would switch without drama if something mildly better appeared. They generate no work item.
Organisations act on work items. Where nothing arrives there is no owner, and where there is no owner there is no action. Nobody decided to ignore this group; nothing ever reached a desk that required a decision about it. That is a harder failure to correct, because there is no meeting in which it visibly occurs.
The Strategic Context
Every feedback instrument has an activation threshold. A complaint requires enough irritation to justify the effort. A testimonial requires enough enthusiasm to justify lending your name to somebody else's marketing. A survey response requires enough of either to justify answering at all.
Three thresholds, all filtering in the same direction: towards the tails. An enterprise's knowledge of its customers is therefore assembled from a sample selected on the very variable it is trying to measure.
This is not a new structural error, only an unfamiliar location for it. An organisation that investigates failures in proportion to the damage they caused learns only about mechanisms that happened to fire under bad conditions, and the case for a standing rule overriding that selection is made elsewhere [Related article: Which Failures Does Your Enterprise Decide Not to Investigate?]. The parallel is exact: there the selection variable is consequence, here it is intensity of feeling. In neither case did the enterprise choose what it learns about. Its instruments chose, and circumstance set the instruments.
What Silence Is Assumed to Mean
Silence is read as satisfaction. It supports at least four other readings, and they carry opposite implications.
Indifference. The relationship is functional and shallow. There is no attachment to lose because none was formed.
Resignation. The customer concluded some time ago that saying something changes nothing. This group looks identical to the contented in every dataset and behaves quite differently under competitive pressure.
Normalisation. The irritation is universal in the category, so it no longer registers as an irritation — including, sometimes, irritations the enterprise is charging for [Related article: Which Frictions Are You Profiting From Tolerating?].
Departure already decided. The customer has chosen to leave and sees no reason to explain.
A second misreading compounds the first: response rate is treated as a sampling inconvenience rather than as evidence. Non-response is not random. It correlates with the disposition under study, so the quiet middle is under-represented in the data about the quiet middle — and a rising response rate is not necessarily good news. People answer when they have something to say.
Reframing the Issue
The question is not how to lift the score. It is which customers currently generate a signal, which do not, and what standing rule governs attention to those who do not.
That reframing moves the problem out of the survey team and into governance, where it can actually be resolved. Leaving a segment unmanaged is a legitimate allocation decision — attention is finite and signal from the middle is expensive to obtain. Leaving it unmanaged without ever deciding to is not a decision at all.
A Difference of Tails Is Not a Description of a Population
The Net Promoter Score, a proprietary loyalty measure associated with Fred Reichheld and held as a registered mark, subtracts the proportion of one tail from the proportion of the other [FACT CHECK REQUIRED]. The middle contributes to the denominator and nothing else.
That is a design choice with a consequence worth stating plainly: two organisations with identical scores can hold opposite distributions. One is polarised, with strong advocates and active detractors; the other is uniformly mild, with almost nobody at either extreme. The index cannot distinguish them, and the response each requires is close to opposite. The first has a service failure. The second has no relationship worth the name.
Two cautions follow. The boundaries defining these categories are stated inconsistently across the accounts in general circulation, so no threshold is asserted here [FACT CHECK REQUIRED]; establish which definition your own instrument uses and whether it has changed, because a silent redefinition breaks a trend line without disturbing the chart. And the index is not useless — it is a headline. The obligation is to publish the distribution beside it.
The Signals You Already Own
Behaviour is a signal, and behaviour does not require anyone to feel strongly.
Purchase frequency and its trend. Order size and breadth of range. Time taken to renew, measured from when the option opened rather than when it closed. Latency in responding to your communications. The share of the relationship conducted through your cheapest channel. None of it depends on volunteering, which is exactly why it describes the middle when nothing else does.
The obstacle is rarely availability. These indicators already sit in systems the enterprise has paid for. They arrive as operational exhaust, owned by functions that do not read them as customer intelligence, and where they are reported at all they are reported deep inside operational packs that no executive reads to the end [Related article: Why Your Best Evidence Sits on Page 22].
Valuing the middle properly then depends on which engine the business runs. In a repeat business the quiet middle is tenure at risk and margin already banked in the forecast. In a referral business it is a referral engine idling — customers who would recommend if asked, and are never asked [Related article: Would This Business Survive If Every Customer Bought Once?].
Decision Framework
Build a signal map first. For each material segment record what signal it generates, who receives it, what triggers an action, and what a reliable signal would cost to obtain.
The governing rule is short: every material segment has a named owner and a defined trigger, or is explicitly declared unmanaged with a review date. Unmanaged is defensible. Unowned is not a position anybody has taken.
Three tests establish where you stand.
The intensity test. What share of last quarter's customer actions were triggered by inbound contact rather than by analysis somebody performed? If it approaches all of them, the enterprise does not have a customer function. It has a complaints desk with a budget.
The non-response test. Who did not answer, and are they different from those who did? Answering requires deliberate stratified outreach to non-respondents — the only way to know whether your feedback describes your customers or your enthusiasts.
The silent-churn test. Of customers lost last year, how many had no support interaction and no survey response in the preceding period? That number is the measured size of the blind spot, and it is usually the figure that makes the argument land in a boardroom.
From Strategy to Execution
Immediately, run the silent-churn test on last year's departures and publish the distribution behind whatever index you report. Both use data you already hold, and either is enough to establish whether the problem is real in your business rather than merely plausible.
Over two or three quarters, build the behavioural instrument and assign the middle to a named executive. The assignment matters more than the instrument: an indicator with no owner becomes another chart, and the enterprise will have paid to observe something it still does not act on. Fund stratified outreach at a scale that gives a defensible read, and accept that this costs more per unit of signal than reading complaints — that is why nobody does it, and pretending otherwise loses the argument at the first budget review.
Over years the durable asset is a base rate. An enterprise holding consistent behavioural definitions for five years can say whether the middle is drifting and by how much, rather than debating whether a two-point score movement means anything. Definitional churn destroys that, which is why the instrument should be dull and stable rather than sophisticated and annually revised.
Signals to Monitor
- The share of churn arriving with no prior contact of any kind, which measures blindness rather than dissatisfaction.
- Response rates by segment and their direction, read as evidence about who is answering rather than as data quality.
- The ratio of customer actions triggered inbound to those triggered by analysis.
- A stable index sitting over a widening distribution — the most misleading pattern in customer reporting.
- Behavioural drift in the middle: falling frequency, narrowing range, migration to the cheapest channel.
- Changes to survey wording, scale or timing, which break comparability silently and are rarely minuted.
Questions for the Leadership Team
- What share of our customers produced any signal at all last quarter, and what do we know about the rest?
- Who owns the customers who said nothing, and what are they expected to do?
- If our index held steady while our distribution polarised, would we know before it showed in revenue?
- Of the customers we lost last year, how many gave us a warning we could have read?
- What would a reliable signal from the middle cost, and what would we do differently if we had one?
- Which of our instruments select on intensity of feeling, and which on behaviour?
Closing Perspective
The score is a difference between two tails. The business is in the middle.
An enterprise that hears only from people who feel strongly has handed its customer agenda to the intensity of other people's emotions — the same delegation, in a different domain, as an organisation that investigates only the failures that happened to be expensive. Neither chose what it would learn about. Circumstance chose, and both mistook the resulting silence for the absence of anything to know.
Correcting it is not primarily a measurement exercise; the instruments are mostly already in the building. What is missing is the decision that a customer who says nothing is still somebody's responsibility, and the willingness to fund finding out what they think — knowing that the answer, when it arrives, will be less dramatic and more expensive to act on than either of the tails.
About the author
Kevin Jogin is Founder & Principal Advisor at EraNorth. Meet the Founder.