Customer State and Attribution in Quality Measurement

Three customers in three states reach the same agent and leave three different scores; the disrupted path arrived that way.

Customer state and attribution in quality measurement concerns a dimension of case mix that quality instruments routinely ignore: the state the customer is in when they make contact. In any service that supports a journey — travel, logistics, healthcare scheduling, home services, insurance — the same request can arrive from a customer who is planning, a customer who is mid-journey, or a customer whose journey is already failing, and the same execution produces very different perceived quality across the three. The disruption case has a further property: the provider usually did not cause the failure — a weather system, a carrier, a supplier — but the provider does own its attribution, because whether the customer blames the weather or the provider is decided by how the recovery goes. Contacts in a disruption state therefore arrive pre-loaded with sentiment that delivery did not create, and scoring them on the same instrument and target as planning contacts measures the weather. This page defines the states, sets out the attribution mechanism and its evidence, and gives the adjustment and routing rules that follow.

Attribution here means where the customer assigns the blame. The recovery of the operation itself — assets, crews, schedules — is at Irregular operations; the closed-loop service-recovery process and the decomposition of experience scores by driver, which that literature also calls attribution, are at Customer Experience Management. The two-axis case-mix framework this page's state dimension belongs to is at Stakes and Complexity: Two Axes of Case Mix; the sentiment instruments are at Sentiment Analysis and CX Signal Integration; the emotional cost to the people handling disruption work is at Emotional Labor in Service Operations.

Three states

• Planning. The customer is arranging something that has not started. Time is available, alternatives exist, and the emotional register is neutral or positive. Execution quality is judged on completeness and convenience.
• In-journey. The customer is inside the thing being arranged — traveling, awaiting a delivery, between appointments. Time is constrained, alternatives are narrowing, and the contact is often a change or a confirmation. Execution is judged on speed and certainty.
• Disruption. Something has failed — a cancellation, a delay, a lost item, a missed connection — and the customer is contacting to recover. Time is critical, alternatives are few, and the emotional register is adverse before the conversation begins. Execution is judged on whether the recovery worked, and the judgment is harsh.

The three states are not three levels of difficulty. A disruption contact may be procedurally simple; a planning contact may be intricate. They are three different relationships between the customer and the service at the moment of contact, and they map onto the stakes axis described at Stakes and Complexity: Two Axes of Case Mix — the state is the stakes axis seen from the customer's side.

Attribution: owning the failure versus owning the blame

In a disruption the provider is rarely the cause. The flight was canceled by the carrier; the road was closed by the weather; the part did not arrive from the supplier. What the provider controls is the recovery, and the recovery is what the customer evaluates. The service-encounter literature established early that the same underlying event — a failure in the delivery system — is remembered as satisfactory or dissatisfactory according to how the employee responded, with roughly a quarter of satisfying encounters in the original study arising from recovery from a failure; the corollary is that a failure badly recovered is attributed to the provider whatever its cause. The provider does not own the failure. It owns the attribution.

Two consequences follow for measurement. First, disruption contacts arrive with negative sentiment already present, generated by the event and not by the delivery; an instrument that scores the conversation without knowing the state records the event as a delivery defect. Second, the score's sensitivity to execution is far higher in the disruption state than in the others: the difference between a good and a poor recovery moves the customer's evaluation more than any difference between a good and a poor planning interaction. The same delivery population, on the same instrument, will show a large gap between its planning and its disruption scores that is not a capability gap.

The evidence on where recovery belongs

Experimental work on customer evaluations of foreign and domestic service employees found that competent agents were rated equally regardless of their labelled origin — the presumed penalty for offshore or outsourced service did not appear when the service was delivered well — and that the penalty appeared specifically where no recovery was offered; where both were incompetent the domestic employee was judged the more harshly, so the effect is not a general penalty on distance. The finding reshapes a common placement argument. The case for keeping customer-facing work with the deepest, most established teams is weak as a general rule — competent delivery is judged as competent wherever it comes from — and the evidence points to recovery moments as the place to look, on findings from vignette experiments mediated by participants' ethnocentric beliefs rather than from operating data. The disruption state is where the placement stakes concentrate, and it is the state most quality instruments cannot see.

The adjustment

Customer state is measurable from fields most service records already hold, and they are the same fields that measure stakes: whether the customer is mid-journey; whether the contact was triggered by an external event rather than initiated by the customer; how close the customer is to a deadline; whether a commitment already exists. A three-state classification built from those fields is coarse and sufficient.

1. Tag every contact with its state from the record, at the time of contact, before any score is computed.
2. Report scores by state before reporting them blended. A population's planning score and its disruption score are two findings, not one.
3. Compare like with like. Two populations are compared on the same state, or on state-adjusted scores, never on raw blends whose state mix differs. This is the case-mix adjustment Sample Size and Detectable Difference in Quality Measurement requires, with state as one of the axes.
4. Set the target by tier, then adjust the instrument by state. The standard the customer bought does not change with the state; the reading of the instrument does, because part of what it reads in a disruption is the event. After-Hours Service as a Distinct Work Type applies the same rule to the work type that lives almost entirely in the disruption state.
5. Weight recovery outcomes separately. Whether the recovery worked — the customer's journey resumed, the loss was made good — is a distinct outcome measure from how the conversation felt, and in the disruption state it is the one that predicts attribution.

The routing consequence

Because attribution is decided in the recovery, disruption-state contacts are the strongest case in a service estate for routing on the high-stakes row of the stakes-and-complexity grid — to a handler with the authority to act now, which for the simple disruption contact means immediacy rather than depth, and for the complex one means the deepest handler available. Routing them by procedural complexity alone sends the simple disruption contact to the widest, slowest-to-act pool, which is the pool least able to recover. The state tag is therefore a routing input as well as a measurement adjustment. Next Generation Routing describes the multi-objective routing that would consume it; its input ladder reaches agent state and interaction value but carries no customer-state term, and the state tag is proposed here as an addition rather than a description of that model.

Failure modes

• Scoring the weather. Disruption contacts are scored on the same instrument and target as planning contacts, and the delivery population is judged for an event it did not cause.
• Reading the state gap as a capability gap. A population handling more disruption work scores lower, and remediation is aimed at training.
• Routing recovery by complexity. The simple disruption contact goes to the pool least able to act at once, and the recovery fails where it is judged hardest.
• Blending states across nodes. Two nodes with different state mixes are compared on raw scores, and the one holding the disruption work loses.
• Ignoring the cost to the handler. A pool that carries the disruption state all day carries the emotional labor described at Emotional Labor in Service Operations, and its attrition is read as a management problem rather than a work-design one.

Maturity Model Position

Tagging contacts by state and reporting scores by state are Level 3 disciplines on the WFM Labs Maturity Model™; the state-adjusted comparison between populations is Level 4, matching where Stakes and Complexity: Two Axes of Case Mix places the axis this state dimension belongs to. The state tag is also a Level 4 routing input, because value-based routing that cannot see the customer's state will place the highest-attribution contacts in a service estate by their procedural weight alone.

See Also

• Stakes and Complexity: Two Axes of Case Mix — the stakes axis, of which state is the customer's-side reading
• After-Hours Service as a Distinct Work Type — the work type whose contacts arrive almost wholly in the disruption state
• Sentiment Analysis and CX Signal Integration — the instruments that read the pre-loaded sentiment
• Sample Size and Detectable Difference in Quality Measurement — case-mix adjustment and detectable difference
• Comparing Delivery Arrangements — the conditions for a fair comparison
• Emotional Labor in Service Operations — the cost of carrying the disruption state
• Irregular operations — recovering the operation itself
• Customer Experience Management — closed-loop service recovery and score attribution by driver