Net Conduct Score, or NCS, gives the human experience of an eligible interaction a structured place in a system that would otherwise record only transactions and business outcomes.
Based only on how this customer interacted with you and others, would you want to serve them again?
An eligible worker answers only after enough direct interaction. Spending, loyalty, prior ratings, customer value, and internal risk labels are not shown.
Definitely yes+10
Probably yes+3
Probably not−3
Definitely not−10
Not enough interaction to sayOutside the score
The weights, aggregation, minimum evidence, recency, confidence, and public presentation remain research candidates rather than production settings.
WHAT IT MEASURES
One worker's willingness to serve one customer again.
NCS represents an eligible direct participant's willingness to serve the customer again, based only on how that customer interacted with people during one verified interaction.
It can include patience, consideration, cooperation, directness, frustration, disagreement, or emotional intensity in context. It does not require cheerfulness, deference, small talk, cultural conformity, or making the work effortless.
WHAT IT EXCLUDES
Nine categories the standard keeps out of NCS.
Spending, profitability, status, loyalty, or lifetime value
Complaint frequency, support use, or cost to serve
Compliance, obedience, agreeableness, or friendliness
Fraud, safety, credit, security, or identity risk
Manager opinion, automated inference, or internal notes
Employee performance or productivity
Public reputation, moral character, or universal rank
Tone, face, sentiment, demographics, or social media
Free-text notes, or any written account of what a person is like
The last two are the ones that matter most in practice. There is no comment box. A rater picks one of five options, and a negative answer adds one category from an approved list. No narrative account of a person is captured, stored, or scored anywhere in the instrument, because a written note asserting what kind of person someone is cannot be corrected by aggregation, never ages out, and follows the person indefinitely. A category can be counted, audited for bias, and appealed.
The last item is the one the design refuses most firmly, so the specification states it as three constitutional requirements rather than an exclusion. A response must be a deliberate act by a person who was present. No automated system may generate, infer, pre-fill, rank, or suggest one, because a worker confirming a suggested answer produces an inferred result recorded as a human judgment. Facial analysis, voice and speech analysis, prosody, sentiment classification, affect detection, and biometric signals are prohibited as inputs, and no response may train a model that infers emotion or conduct about anyone.
A real interaction is verified through an approved source.
02
The worker directly participated and had enough exposure to answer.
03
The role, response window, notice, and capture method match the approved context.
04
The interaction was not selected to produce a favorable or unfavorable response.
05
The worker can answer privately. Whether abstaining is safe in practice, especially in a small team, is unresolved.
06
Prior CQ, loyalty, value, risk, fraud, and customer labels cannot influence the response.
RIGHTS ON BOTH SIDES
Privacy for workers. Review for consumers.
Worker protection
Answering is never mandatory, and no business may require a response or make one a condition of the role. No manager scripts, quotas, approval, performance use, retaliation, or individual-response access. The standard requires that capture and any required training, review, or appeal participation be paid as work in every jurisdiction, not only where local law compels it, and no wage-hour analysis supports that requirement yet.
Two limits remain unresolved. Whether declining feels free when an employer scheduled the task is a labor question rather than a design one. And a minimum number of eligible responders per location is required before capture can operate there at all, because in a very small team an abstention is itself identifying. That minimum does not exist yet.
Consumer protection
Consumers can correct source facts, question eligibility, report bias or selective sampling, add context, review the calculation, and escalate material harm. None of that requires learning the worker's identity or entering a personal dispute with them.
BEYOND CQ
NCS may have uses beyond customer relationships.
Agencies and clients, businesses and vendors, event participants, community hosts, and collaborators may find the question useful. None can inherit the consumer rules automatically.
Power, dependency, retaliation, collective activity, employment law, and access differ by context. Putting the NCS question into an employee review would not make that system aligned with Open Conduct.
HOW IT FITS
How the two inputs combine.
Net Conduct Score records the employee's experience. Verified Conduct Index uses approved business records to evaluate eligible responsibilities the customer accepted and could control.
When both inputs independently qualify, Conduct Quotient may combine them into one private score for a named context. Each input must pass its own gates first, so neither can substitute for the other. The synthesis formula then caps how far a strong input can lift a weak one.
WHERE IT COMES FROM
NCS reverses the Net Promoter Score question.
Net Promoter Score asks a customer how they feel about a company. NCS asks a company's people how they feel about a customer. The shape is deliberately the same: one item, a small forced-choice set, and directional weights. Naming the lineage matters because it imports the criticism along with the format.
Single-item instruments carry known problems. Reliability cannot be estimated from one item the way it can from a scale. NPS in particular has been criticized for arbitrary category boundaries, for discarding information when responses are collapsed, and for producing differences that look meaningful and are not. Every one of those criticisms applies here and none is answered yet. The response weights, the cut points, and the decision to use a single item are all research candidates rather than settled choices.
RESEARCH AGENDA
Seven research questions that would settle the design.
Does the question measure willingness to serve again for the intended reason?
Do the four response labels communicate strength consistently across cultures, languages, roles, and contexts?
Does removing the midpoint clarify direction or force a judgment that is not there?
Is the abstention option understood and safe to use?
What history is sufficient before any summary becomes available?
Can workers respond without manager pressure, reidentification, or performance fear?
Can consumers understand, challenge, and recover from an incorrect or biased record?
HELP IMPROVE NCS
We need measurement and labor experts.
A scenario the question handles badly, a cultural or labor concern, a research method, a protection worth adding, or simply better wording would each move this forward.