Why Good Incentives Still Fail Without Human Taste

The hidden trap: fixing the wrong thing

What if the biggest reason businesses, policies, and teams fail is not that they are badly designed, but that they are designed for the wrong measure of success?

That sounds obvious until you notice how often intelligent people still fall for it. We build metrics to make reality legible, then we accidentally start serving the metric instead of the reality. We reward what is easy to count, what looks impressive in public, what scales neatly in a presentation, and what makes a dashboard turn green. But the things that actually create value are often quieter, smaller, slower, and harder to prove.

This is where many well intended solutions become their own sabotage. A policy that reduces one visible problem may create three invisible ones. A business that optimizes for growth may lose its craft. A team that chases recognition may stop doing the work that matters. The real tension is not between good intentions and bad outcomes. It is between what is measurable and what is meaningful.

The deepest failures are often not failures of effort. They are failures of translation, when a human goal gets converted into a crude proxy and then mistaken for the goal itself.

The cobra bounty story is a perfect metaphor for this. A reward intended to reduce a problem instead made the problem more profitable. That is not just a policy error. It is a universal warning: whenever a system pays for the appearance of progress, people will eventually learn how to manufacture that appearance.

---

The cobra effect is everywhere, not just in government

Most people think perverse incentives only appear in bureaucracies or public policy. In reality, they are one of the most common forces in modern life. They show up anywhere we confuse evidence of value with value itself.

Think about a company that rewards employees for visibility. Soon people learn to write longer emails, speak more in meetings, and make their work public even when silence would be more productive. Think about a startup that prizes headline growth. It may acquire users who churn, attract attention that does not convert, and fund initiatives that create buzz but not durable value. Think about schools that test for memorization. Students optimize for test performance and lose the habit of thinking.

The same pattern appears in ordinary life. If you praise a child only for results, they may start hiding mistakes. If you compliment a friend only for being productive, they may feel guilty resting. If you use social media likes as a signal of worth, you will begin to edit your identity for the crowd.

The cobra effect is not merely a cautionary tale about bad incentives. It is a reminder that systems are adaptive. People are not passive recipients of your rules. They are pattern seekers. They will discover the shortest path to the reward, even when that path bends around your original intention.

This is why surface level solutions are so dangerous. They often target the most visible symptom, then leave the deeper structure untouched. A city can crack down on visible disorder without reducing desperation. A company can require more reporting without improving decision quality. A founder can chase press without building product love. In each case, the system gets better at looking solved while remaining unsolved.

The lesson is unsettling: if you only inspect outcomes at the level of appearance, you will encourage people to optimize for appearance.

---

The overlooked advantage: small, quiet, and human

Now consider the opposite impulse. Not the impulse to engineer broad incentives, but the instinct to respect craft, context, and the person doing the work.

A small restaurant can beat giant chains not because it has more capital, but because it has a sharper feel for what customers actually care about. It can obsess over seasoning, warmth, timing, and memory. It can learn from one repeat customer in a way a spreadsheet cannot. It can turn a simple dish into a ritual. The point is not that small is always better. The point is that scale often hides signal.

This is why there is something deeply important in the idea that, in the end, it matters less what you do than who does it. That sounds almost too simple, but it captures a truth many organizations forget: execution is not a mechanical property. It is a human one.

Two businesses can sell the same product in the same market and produce radically different outcomes. One treats the work as a checklist. The other treats it as a craft. One optimizes for looking legitimate. The other optimizes for being excellent. One asks, “How do we appear competitive?” The other asks, “What would make this genuinely good?”

This is why even in saturated markets, success is still possible. A crowded field is not a death sentence if you bring a distinct point of view, better taste, deeper care, or stronger discipline. The category does not determine destiny. The quality of attention does.

In many industries, the real moat is not the business model. It is the person or team who refuses to do mediocre work even when mediocre work is rewarded.

That is the bridge between the two ideas. Incentives shape behavior, yes, but human judgment shapes the interpretation of those incentives. A weak operator will exploit a metric. A strong operator will use the metric only as a clue, never as the destination.

---

A better mental model: proxies, craft, and externalities

To connect these ideas more deeply, it helps to use a simple framework with three layers:

1. Proxy layer: what can be counted, reported, or rewarded.
2. Craft layer: what actually creates durable value.
3. Externality layer: what gets pushed onto others when the proxy and craft diverge.

Most failures begin when an organization spends too much time in the proxy layer. It chooses a number because the number is available, then treats optimization of that number as a strategy. But the proxy layer is not reality. It is a translation of reality, and translations can distort.

The craft layer is where real work happens. It includes taste, timing, empathy, judgment, and care. These are not soft extras. They are often the very mechanisms that create lasting differentiation. They are hard to measure precisely, which is exactly why they are easy to neglect and easy to underinvest in.

The externality layer is where the damage accumulates. If a business chases growth at any cost, it may create burnout, customer disappointment, or reputational fragility. If a city responds to a problem with a simplistic crackdown, it may relocate the problem rather than solve it. If a restaurant scales without protecting its standards, it may become famous for something it no longer does well.

This framework explains why some interventions work better than others. Safe injection sites reduce harm not by moral theater, but by acknowledging actual behavior and reducing collateral damage. Sex education reduces teen pregnancy not by pretending desire does not exist, but by equipping people to act with knowledge. These solutions work because they respect the structure of the problem rather than merely the public optics of the problem.

That same principle applies in business. The question is not, “What policy sounds good?” It is, “What behavior will this policy produce?” More importantly, it is, “What behavior will this policy produce once smart, self interested people adapt to it?”

When you ask that second question honestly, you become harder to fool.

---

Why human taste beats generic optimization

There is a reason some founders and operators seem to succeed in places where others fail. They are not always smarter in the abstract. They are often better at tasteful judgment.

Taste is the ability to distinguish between what is merely acceptable and what is truly resonant. It is the thing that notices when a product feels alive instead of functional, when a service feels cared for instead of processed, when a menu feels intentional instead of assembled. Taste cannot be fully automated because it is not just an algorithm for maximizing a metric. It is a trained sensitivity to quality.

This matters because incentives alone do not guarantee excellence. They can push people toward effort, but not necessarily toward the right kind of effort. A team can be highly motivated and still make bad products if the reward structure favors speed over integrity. A founder can be ambitious and still build something forgettable if ambition is disconnected from taste.

The most durable businesses usually combine two things that are often separated:

• Disciplined incentives, so the system does not drift into chaos.
• Human taste, so the system does not drift into mediocrity.

Without incentives, good intentions dissolve. Without taste, incentives become hollow. One protects against laziness. The other protects against cheapness.

This is why some of the best small businesses feel so alive. They are not optimizing to look impressive. They are optimizing to feel right. Customers sense the difference immediately, even if they cannot name it. That is not magic. That is the result of someone close enough to the work to notice what matters and stubborn enough to protect it.

A chain can copy ingredients. It cannot easily copy conviction.

---

The real question: what are you teaching the system to value?

At the center of all this is a more demanding question than “What should we do?” It is “What will this teach people to value?”

Every policy, metric, and reward sends a lesson. Every lesson shapes attention. And attention shapes reality.

If you reward only scale, people will chase scale. If you reward only visibility, people will chase visibility. If you reward only speed, people will rush. If you reward only compliance, people will imitate compliance. The danger is not that people are evil. The danger is that people are intelligent. They will learn whatever your system is truly teaching.

That is why the best leaders, founders, and policymakers act like gardeners rather than engineers. Engineers assume a direct causal chain. Gardeners know that inputs interact, that growth is uneven, and that overcorrecting one problem can create another. They pay attention to the whole ecology. They ask not only what grows, but what crowds out what else.

A good system is not one that eliminates all incentives. That is impossible. A good system is one that aligns incentives with the deeper craft of the work, while leaving room for human judgment to intervene when the proxy starts to lie.

This is especially important in places where the work is fundamentally relational. Restaurants, schools, hospitals, product teams, customer support, local communities: these are not assembly lines. They are living systems. If you reduce them to dashboards alone, you will miss the texture that makes them work.

And yet the answer is not to reject metrics entirely. The answer is to treat them as instruments, not idols.

Metrics should tell you where to look. They should never be allowed to tell you what is real.

---

Key Takeaways

1. Ask what a metric teaches, not just what it measures. Every reward system trains behavior. Before adopting a KPI, ask what kind of gaming it invites.

2. Protect the craft layer. The most valuable work is often the hardest to count: judgment, timing, care, and taste. Build room for those qualities to matter.

3. Watch for proxy drift. If a team starts optimizing the signal of success instead of success itself, the system is already failing, even if the dashboards look healthy.

4. Design for adaptation, not compliance. Assume people will learn the rulebook quickly. Build systems that remain healthy after smart people adapt to them.

5. Remember that people, not categories, create outcomes. Even in crowded markets, exceptional execution and genuine taste can beat incumbency.

---

The uncomfortable but liberating conclusion

The deepest mistake in business and policy is not over ambition. It is confusion about what is actually being optimized.

We are tempted to believe that the right structure will automatically produce the right result. But structure is only half the story. The other half is the human being inside it: the operator with taste, the founder with conviction, the manager with judgment, the policymaker with restraint. Systems matter enormously, but systems do not interpret themselves. People do.

That is why the most powerful combination is not scale plus incentives. It is incentives plus human discernment. Incentives keep reality honest. Discernment keeps the system from becoming stupid.

So the next time someone proposes a simple fix, ask a harder question. Not, “Will this solve the problem?” but, “What new behavior will this create, and who will be smart enough to exploit it?” Then ask the second question: “If the system drifts, who has enough taste to notice?”

The future belongs less to the organizations that can measure the most, and more to the ones that can still tell the difference between a number and the thing the number was supposed to represent.