When Good Metrics Go Bad: The Hidden Economy of Incentives in Organizations

The most dangerous numbers in an organization are the ones that look obviously useful

What if the fastest way to make a team worse is to measure exactly what seems most reasonable? That is the uncomfortable reality of incentives: the moment you turn a goal into a metric, people start optimizing for the metric, not the goal. Sometimes the result is harmless nonsense. Sometimes it is a quiet disaster. And sometimes, as with the classic cobra problem, the attempt to solve an issue ends up manufacturing more of it.

This is not just a story about public policy or quirky historical mistakes. It is a story about how organizations, managers, and teams accidentally train people to game the system. It is also a story about leadership, because the best one on ones are not status updates or calendar rituals. They are places where hidden incentives surface before they harden into culture.

The deeper question is this: how do you create systems that reward the thing you actually want, not the thing that is easiest to count?

---

Every metric is a spotlight, and every spotlight casts a shadow

A common managerial instinct is to make the invisible visible. That instinct is often right. If no one can see what matters, it is hard to improve it. But the moment you select a proxy, you change the game. People do not respond to intent alone. They respond to what is noticed, rewarded, and repeated.

That is why organizations end up valuing things like size, busyness, or press coverage even when those are not the real point. A team may learn that looking large matters more than being effective. A contributor may learn that visible work gets credit while essential but quiet work goes unrecognized. A company may learn that public attention is treated as proof of importance, even though many truly important jobs are invisible until they fail.

This is the structure behind the cobra effect: the solution becomes a new target, and the target becomes the problem. If you pay people to produce cobras, you may get cobra breeding. If you pay teams to maximize output without regard to quality, you may get bloated output and fragile systems. If you reward managers for keeping calendars full, you may get administration masquerading as leadership.

The key insight is not that metrics are bad. It is that metrics are moral hazards when they become too legible. People begin to serve the measurement, because the measurement is what the system can see.

What gets measured gets managed, but what gets overmeasured gets distorted.

---

The real problem is not incentives, but incentive blindness

Many people talk about incentives as if they are obvious levers. In reality, the hardest incentives to deal with are the ones nobody believes are there. Most teams can identify the obvious rewards: promotions, bonuses, praise, and authority. Fewer people notice the quieter rewards: avoiding blame, appearing busy, looking decisive, staying in good standing, or never creating uncomfortable work for your manager.

This matters because people optimize under uncertainty. If a person is not sure what truly matters, they will usually infer it from what is repeatedly inspected. If their manager asks about shipping dates every week but never asks about maintainability, the manager has effectively declared one more important than the other. If one on ones are used mainly to surface blockers that the manager can remove, people will learn to bring only blocker-shaped problems. If they become performance theater, they will hide weakness and present polished narratives.

That is why one on ones are such a revealing leadership tool. At their best, they are not about extracting updates. They are about making incentives speak out loud. They ask questions like: What are you optimizing for right now? What is the system rewarding that it should not? What important work is getting ignored because it is hard to see?

A great manager uses one on ones to discover what the team believes is safe, valued, and worth doing. That is often very different from what the org chart says. If people only talk about what is easy to defend, the manager is hearing the shape of the incentive landscape, not the shape of reality.

A useful framework: the three layers of incentives

Most incentive failures happen across three layers:

1. Declared incentives: what the organization says it wants.
2. Visible incentives: what gets rewarded, praised, or reviewed.
3. Behavioral incentives: what people actually do to stay successful and safe.

When these layers align, systems feel smooth. When they diverge, culture becomes strategic camouflage. People learn to say one thing, measure another, and do a third.

For example, a company may declare that it values craftsmanship, but visibly reward speed and output. The behavioral incentive becomes shipping fast, even at the cost of quality. Or a manager may declare that honest feedback is welcome, while visibly punishing bad news. The behavioral incentive becomes selective reporting and strategic optimism.

Once you see this, the cobra effect stops looking like a weird exception. It becomes the default risk of every poorly designed system.

---

The best interventions often work by changing the environment, not by commanding better behavior

One reason the cobra effect is so seductive is that it appeals to the most common managerial fantasy: if people would just behave better, the problem would go away. But in many cases, durable solutions do not come from exhortation. They come from redesigning the conditions under which people make choices.

That is why interventions such as safe injection sites and sex education are so important as examples. They do not merely ask people to act responsibly. They alter the surrounding conditions in ways that reduce harm. They address the externalities, the hidden costs that flow outward when systems are ignored or moralized instead of managed.

This is a profound lesson for organizations. If your people are producing the wrong thing, the answer is rarely to tell them to care more. More often, the answer is to change the environment so that the wrong thing is no longer rewarded.

Consider a few workplace analogies:

• If engineers are rewarded only for feature output, they may accumulate technical debt.
• If managers are rewarded only for team growth, they may hire too quickly.
• If salespeople are rewarded only for closed deals, they may oversell and create churn.
• If leaders are rewarded only for visible presence, they may mistake attendance for impact.

The smart move is not to install more pressure. It is to align the surrounding system with the intended outcome.

The highest leverage leadership does not demand better motives. It makes better behavior easier and bad behavior less profitable.

This is why the distinction between problem solving and incentive design matters. Problem solving asks, “How do we fix this case?” Incentive design asks, “What will happen every time we create this rule?” One is local. The other is systemic.

---

One on ones are where incentive design becomes personal

A lot of leaders treat one on ones as a soft management ritual. They ask how things are going, maybe discuss priorities, then move on. But one on ones are actually one of the few places where you can observe the lived experience of the incentive structure before it turns into attrition, politics, or dysfunction.

If a team member never raises risk, that may not mean there is no risk. It may mean risk is costly to mention. If someone only brings positive updates, it may not mean they are thriving. It may mean vulnerability is punished. If a person spends the meeting reassuring you, they may be optimizing for manager comfort rather than shared clarity.

A manager who understands incentives listens for signals beneath the words:

• What does this person avoid saying?
• What topics get minimized?
• Where do they speak with energy versus caution?
• Which problems are framed as technical, when they may actually be structural?
• What work is invisible, unrewarded, or emotionally expensive?

These questions matter because many organizational failures do not begin with malice. They begin with silent adaptation. People learn what to emphasize and what to hide. Then those adaptations become habits, and the habits become norms.

Think of a team as a pond. The meeting is the surface. Incentives are the currents below it. If you only look at the surface, you may think the water is still. But the fish are moving according to forces you cannot see unless you know where to look.

The best one on ones are therefore not a tool for control. They are a sensor for distortion. They tell you whether your system is producing candor or compliance, ownership or theater, learning or caution.

---

A better mental model: optimize for the second order effect

Most incentive failures happen because people focus on the first order effect and ignore the second order effect.

The first order effect is the thing the rule seems to solve right away. Paying for cobras reduces cobras. Asking for more output increases output. Tracking attendance improves visibility. Rewarding speed improves velocity.

The second order effect is what happens after people adapt. Cobra breeding appears. Output quality falls. Meetings become performance. Speed turns into rework.

This leads to a practical leadership rule: never design for the direct effect alone. Ask what the smartest person in the system will do once they notice the rule.

A few diagnostic questions help:

1. What will people optimize if they want to look good?
2. What will people optimize if they want to stay safe?
3. What will people optimize if they want to avoid blame?
4. What important thing is hard to count, and therefore likely to be neglected?
5. What behavior would be rational even if it is not desirable?

These questions expose the hidden tradeoffs that most systems ignore. They also reveal whether your intended values are actually embedded in daily incentives or merely displayed on a wall.

A team that values deep work but constantly interrupts people for urgent pings will not get deep work. A company that values innovation but punishes failed experiments will not get innovation. A manager who says they want honesty but reacts badly to bad news will get delay, editing, and politeness.

The principle is simple, though not easy: people will do what the system makes sensible.

---

Key Takeaways

• Assume every reward creates a shadow incentive. Before adopting a metric or rule, ask what people might do to win it without producing the real outcome.
• Treat one on ones as an incentive audit. Listen for what people are afraid to say, what they overemphasize, and what work remains invisible.
• Prefer environmental fixes to motivational speeches. If behavior is broken, redesign the context so the right behavior becomes easier and the wrong behavior less attractive.
• Track second order effects, not just first order wins. Ask how the system will behave after people adapt to the new rule.
• Reward what is hard to see but essential. Quiet reliability, maintainability, mentorship, truth telling, and prevention often matter more than flashy output.

---

The highest form of leadership is incentive humility

The deepest lesson here is not that people are cynical. It is that systems are more creative than we are. They discover loopholes, exploit ambiguity, and convert noble intentions into strategic behavior. That is why the most responsible leaders are not those who believe they can command virtue. They are those who understand that every structure teaches people what to value.

In that sense, managing incentives is less like pushing buttons and more like gardening. You do not force growth by shouting at the plants. You shape the soil, light, water, and spacing so the right things can grow and the wrong things struggle to survive.

The cobra effect is a warning against naive fixes. But it is also an invitation to think more carefully about how organizations really work. The goal is not to eliminate incentives. The goal is to make them legible, honest, and aligned with the outcome that actually matters.

If you do that well, one on ones become more than meetings. They become the place where reality quietly corrects the organization before the organization corrects itself in the worst possible way.

That may be the most important management skill of all: not getting clever about control, but getting humble about causality.