What We Reward
Is Who We Become

Imagine a customer asking why a refund has not arrived. The agent follows every prescribed step, confirms that the request was logged, closes the case quickly and receives a strong quality score. Two days later the customer calls again. The refund has stalled between teams, but nobody who handled the first contact saw that handover fail. We will use this imagined case to follow the book’s argument. Each interaction may look successful on its own; the customer’s journey tells a different story.

The book opens with a QA joke: if you want a good score, pay closer attention to the rubric than to the problem. People laugh because they recognise the situation. Quality assurance systems were often designed for good reasons: consistency, customer protection, compliance and control across large volumes of work. The trouble begins when the work changes faster than the scorecard. A person may be asked to investigate, weigh risk, prevent recurrence and exercise professional judgement while being evaluated most decisively on whether the call was short and the required phrases were spoken.

People notice which signals carry consequences. If a careful investigation increases handling time, and handling time determines standing, the system has given an instruction regardless of what the leadership speech says. Following the rubric becomes the rational route to a good review. The refund is still missing. The scorecard, however, is satisfied. Nobody has to intend the distortion for it to occur.

That is the meaning of the title. Over time, a reward system shapes what people practise, what they avoid and what kind of professional they can afford to become. To understand a culture, examine the behaviour that earns approval on an ordinary Tuesday.

Speed, volume and closure are useful operational signals. Leaders need to know whether demand is rising, customers are waiting or staffing is stretched. The book does not ask an organisation to discard these measures. It asks which measures receive the most authority when they conflict with other evidence.

For the refund query, the short call and completed checklist describe something real. They do not establish that the money reached the customer. Output tells us what the team did. Outcome asks what happened as a result, including what happened later. Both belong in view, with their different limitations understood. If resolution time falls across the team while repeat contacts rise, the conflict between those signals deserves investigation.

Valuable work can produce less visible activity. Finding the cause of an error, correcting guidance or escalating a risk may take longer today and prevent future cases. These contributions are difficult to attribute and easy to crowd out when immediate throughput dominates.

Recurring trouble is therefore worth following across queues and time. A closed case returning through another channel is not automatically proof of a poor decision. Some recurrence is unavoidable. But a cluster of similar returns can reveal a fragile fix, a broken handover or a system that pays people to close work before it is understood. The useful question is whether recurrence is investigated and owned, rather than treated as a fresh, unrelated case each time.

Prevention creates a strange evidence problem: when it works, the future cases never arrive. That absence may be a clue, but it is difficult to attribute and should never become a simple individual target. The book asks leaders to investigate it cautiously, alongside the cases that do return, rather than claiming every reduction as proof of one person’s impact.

Good judgement can appear in a decision to follow a strict rule. Data protection, safety and legal obligations often leave little room for discretion. The question is whether the person understood the rule’s purpose and applied it appropriately to the situation. Compliance can be essential while still being incomplete as an account of quality.

The book makes judgement more concrete by looking at decisions. In the refund case, could the agent distinguish a logged request from an approved payment? Did they recognise that the handover status was unknown, investigate within their access and authority, and give the next team the context needed to act? Those traces of reasoning can be discussed and compared in real cases. The agent should also be able to identify the point where further investigation would exceed their authority and a clear escalation is needed.

An outcome alone cannot settle whether the decision was sound. A careful choice may still end badly because information was missing or another part of the system failed. A weak choice may succeed through luck. Nor does a polished explanation prove good thinking: once people know a narrative is scored, they learn to perform the approved language. The closer a measure gets to professional judgement, the more care its use demands.

The book’s answer is selective examination and calibration. Reviewers look together at comparable, complex cases, explain where their interpretations differ, and develop shared standards for strong, acceptable and weak reasoning in context. Some disagreement is legitimate. Consistent, unexplained divergence is a problem to govern. Fairness comes from explainable decisions, transparent standards and a safe way to challenge an interpretation, rather than from making every judgement look like the same number.

A metric can begin as a sensible response to a real problem and lose its meaning as tools, customer expectations and behaviour change. The chart still works. Its explanation of reality does not.

Watch for performance scores rising while the customer’s experience stands still. Notice when reviewers need more and more caveats to defend a number, when everyone’s case notes start sounding rehearsed, or when questioning the metric itself becomes politically uncomfortable. Those are signs that the measure deserves investigation. They are not an invitation to accuse staff of gaming the system: people are often adapting reasonably to the incentives placed in front of them.

Responsible measurement includes an owner, a reason for using each signal, a review date and conditions for changing or retiring it. Retirement requires care because an old metric may still help with capacity planning, comparisons or risk control. The leader who can introduce a measure must also have the authority and responsibility to decide when it no longer deserves the same weight. A deliberate transition preserves accountability while making room for a more truthful signal.

Prevention, knowledge, collaboration and recognition form the book’s middle argument. They are often discussed as separate programmes, but the scorecard connects them. Each asks whether people have time and permission to make tomorrow’s work better, or whether today’s visible output consumes all available attention.

Knowledge is a good example. After repeated refund queries, an agent may learn that “request logged” is often mistaken for “payment approved”. That insight is useful only if the guidance can be corrected and the handover checked. An organisation may own hundreds of articles and still fail to learn if no one has time or authority to update them. Blaming frontline staff for the gap misses the governance and capacity choices that created it.

Collaboration has a similar test. The service team can explain the repeated refund contacts; the payments team can identify the stalled approval. A meeting helps only if someone can decide how the handover will change and who owns it. When teams have different targets, an issue may circulate through forums without that decision. Frontline insight should inform the fix without making the agent carry responsibility they cannot exercise.

Recognition also teaches. A rescue makes a good story. If the organisation needs a fresh hero every Tuesday, it may need a better system. Repeated hero stories can teach people that visible rescue matters more than quiet prevention. Praising a successful outcome without examining the decision may reward luck or unnecessary risk. Thoughtful recognition notices the person whose careful work stopped a problem recurring, acknowledges the conditions that made the contribution possible, and resists turning judgement into a competition for the most impressive story.

That makes recognition different from a points scheme. It should explain a contribution without giving people a predictable script for earning applause. When a cross-team fix depended on several people, the recognition should reflect the collective work rather than force a single hero into the story.

Leaders must give this work time and decision rights, even when prevention looks less dramatic than another rescue.

AI can draft a polished answer to the refund question in seconds. If the system rewards a fast, correctly phrased closure, that answer may look excellent even though it has not checked whether the payment was approved. AI can speed up drafting, searching and recommending; it can also scale an organisation’s existing preferences. The consequences of weak judgement may then spread more quickly.

Human oversight must mean more than placing a person after a machine in a workflow. The person needs to recognise when an output lacks context, when a confident answer exceeds the evidence, when a case requires escalation and when automation should not make the decision. These are capabilities that require time, training, authority and review. “A human checked it” says little about whether that human was able to exercise judgement.

As tools and policies change, past competence can become outdated while confidence remains high. The book calls attention to learning velocity: how readily people update their thinking when conditions change. It appears in better decisions over time, not merely in completed training modules. Making learning another loud target would encourage people to display it rather than do it. The organisation needs capacity for learning and a way to notice when it changes the work.

If the existing scorecard is incomplete, immediate replacement may sound decisive. It can also make people unsure how to protect their standing while new standards are still being interpreted. The book proposes adding before subtracting: keep the operational system in place while a small sample of the same work is examined through a second lens. For the refund case, one review records whether the required steps were followed; another asks what the agent could reasonably know, whether the handover was investigated, and whether the resolution held. The case is handled once. The parallel activity is in its interpretation, for learning rather than immediate performance consequences.

This gives reviewers time to test whether the new lens reveals sound reasoning, prevention and appropriate escalation, whether they can calibrate disagreements, and whether the lens creates new performance theatre. The pilot needs a defined question, boundaries, protection for participants and explicit authority to stop. Its purpose is trustworthy evidence. A persuasive success story is not enough. Equally, parallel measurement cannot become permanent clutter; leaders must eventually decide what to keep, revise or retire.

Even a modest pilot needs the basics: reliable case records, a way to connect repeat work over time and clear non-negotiable safety or compliance rules. Without those foundations, an apparently sophisticated judgement measure cannot tell leaders whether a resolution held. More data will not repair an unclear decision. Before adding a signal, ask what decision it will support, what behaviour it may encourage or suppress, and who will notice when it stops helping.

Leadership readiness is a real constraint. Calibrating difficult cases takes attention and skill. People need a protected way to say that interpretations are unfair, capacity is missing or the pilot should pause. Sometimes a team is ready to learn while the organisation is not ready to change evaluations at scale. Pausing with a path to build capability is more responsible than transferring the risk of an unfinished system onto staff.

There is a difference between leaders who lack the time or skill and leaders who avoid an uncomfortable result. The first need support and a slower sequence. The second require governance and honest escalation. Treating both as ordinary “resistance to change” hides who is carrying the risk.

The findings also have to travel upward honestly. Senior leaders need a clear account of what the evidence suggests, how confident the organisation is, what remains uncertain and what decision is required. Compressing complexity into a reassuring number may make a presentation easier while making the decision worse. Leaders set the tone by allowing “we do not know yet” to be said without punishment.

An executive deciding what to try next needs a different account from an auditor asking whether a requirement was met. The book warns against forcing both purposes into one apparently precise number. The people translating the evidence need authority and protection to preserve that distinction.