Incentives
The most powerful — and most misunderstood — force in any organization. People don't do what you tell them; over time, they do what they're rewarded for. Get this wrong and no mission statement, pep talk, or org chart will save you.
🎯 By the end you'll understand
Why incentives quietly overrule instructions, mission, and logic; how rewarding a proxy produces the wrong behavior (the cobra effect and Goodhart's law); the principal–agent problem; and how to design incentives that are hard to game — plus the honest limits of trying to incentivize everything.
KEY INSIGHTThe one idea to take away
“Show me the incentive and I'll show you the outcome.” — Charlie Munger
You can announce a strategy, write a values deck, and ask people nicely a hundred times. But standing behind every one of those words is a quieter, stronger signal: what actually gets rewarded around here. When the two disagree, the incentive wins — not because people are cynical, but because incentives shape behavior continuously, while your words shaped it once. If you want to predict what an organization will do next year, don't read its mission. Read its comp plan, its promotion criteria, and what gets someone praised in a meeting.
Incentives are a system input, not a moral one. People aren't being bad when they follow their incentives — they're being rational. So when behavior frustrates you, the first question isn't “why are they like this?” It's “what is this behavior being rewarded for?” Change the reward and the behavior follows. Exhort without changing the reward and nothing moves.
DEFINITIONWhat an incentive really is
An incentive is anything that makes a behavior more or less attractive to the person choosing it. The classic mistake is to read that as money. Money is one incentive — often not the strongest. The full list of what people actually optimize for is much richer, and you met most of it back in Lesson 2.4 (Motivation):
- Status & recognition — being seen as the person who did the impressive thing.
- Autonomy — control over how, when, and what you work on.
- Security — not getting fired, not looking incompetent, keeping the team intact.
- Promotion & advancement — the title, the scope, the path upward.
- Avoiding blame — frequently the single most powerful incentive in a large organization, and the one no one designs on purpose.
This matters because most managers think they have one incentive lever (pay) when they're sitting on a console of ten. A public thank-you, who gets the interesting project, whose name is on the launch, who gets thrown under the bus when something breaks — every one of these teaches people what to do next. You are always running an incentive system. The only question is whether you designed it or it designed itself.
A reward or penalty — monetary or not — that changes how attractive a behavior is to the person deciding. Includes status, recognition, autonomy, security, promotion, and the avoidance of blame, not just compensation. What gets rewarded gets repeated.
CORE CONCEPTThe cobra effect: rewarding the proxy
Here's where it gets dangerous. You almost never can measure the thing you actually want. You want “a great product,” but you can only measure tickets closed, lines of code, revenue, or NPS. So you reward the proxy — the measurable stand-in. And the moment you do, people start optimizing the proxy instead of the goal.
The folk story gives this its name. Colonial administrators in Delhi, the tale goes, wanted fewer cobras, so they paid a bounty per dead cobra. Soon people were breeding cobras to collect bounties. When the program was cancelled, the worthless snakes were released — and the cobra population rose. Whether the story is literally true or not, the pattern is real and everywhere. This is the cobra effect: a reward for a proxy produces exactly the behavior you didn't want.
| You reward this proxy… | …you get this gaming |
|---|---|
| Lines of code written | Bloated, verbose code; nobody deletes anything |
| Bugs fixed (paid bug bounty, internal) | Bugs quietly introduced so they can be “found” |
| Sales bookings against a quota | Sandbagging deals into next quarter; discounting to close junk |
| Tickets closed per day | Closing tickets fast without solving them; reopening churn |
| Calls handled per hour (support) | Hanging up on hard customers to protect the average |
The deep version of this is Goodhart's law: “When a measure becomes a target, it ceases to be a good measure.” A metric is useful precisely because it's a side-effect of good work. The instant you reward it directly, people decouple the metric from the work and optimize the number. The proxy and the goal drift apart — and you've paid for the drift.
Goodhart's law: when a measure becomes a target, it stops measuring what it used to. A perverse incentive (a.k.a. the cobra effect) is the result — a reward structure that produces the opposite of the intended outcome because it rewards a gameable proxy rather than the real goal.
This is also why incentives are the lever for changing system behavior, the idea from Lesson 1.6 (Systems Thinking). An organization is a system; incentives are one of its strongest feedback loops. To change what a system produces, you change its loops — not the loudness of your requests. “Please write better code” is exhortation. “Code quality is now a promotion criterion” is an incentive. Only one of them reliably moves the system.
SEE ITBreak an incentive, then fix it
You run a support desk of four agents. This month 120 customers will come to it with a problem, and you set one thing: what the team's bonus pays for. Pick a rule and let the month run. The board at the end of the belt lights a square for every bonus paid. When the month is over, count how many customers were really helped, then run another month under a different rule. Your goal is the most customers helped.
Notice what happened under tickets closed: nobody disobeyed you. The agents closed tickets fast without fixing them, the same customers came back angrier, and every return was one more ticket to close. (A real desk adds two tricks the scene leaves out: splitting one problem into several tickets, and steering clear of the hard cases.) A closed ticket is a proxy for a solved problem, and under a bonus the two came apart: your number went up while the customers got worse.
The rule that worked did not ask the team to care more. It paid for two numbers that check each other: resolved on the first contact, and the customer satisfied with that ticket. Closing a ticket unsolved now fails both, so the cheapest way to earn the bonus is to do the job. It still left the hardest case for last, which is why no rule can be set once and forgotten.
The same thing happens away from the support desk. Pick another team's metric, read what it buys, and see what a better-designed incentive would pay for instead.
Notice the last option behaves differently. Rewarding the outcome you actually want — and using more than one metric so no single number can be gamed in isolation — is the closest thing to a defence. It doesn't make gaming impossible. It makes the cheapest way to win the reward roughly the same as doing the real job.
CORE CONCEPTAligned vs misaligned: the principal–agent problem
Economists have a name for the core tension: the principal–agent problem. A principal (you, the founder; the board; the customer) hires an agent (an employee, a manager, a contractor, a salesperson) to act on their behalf. But the agent has their own goals — and they will optimize their incentives, not your goals, whenever the two diverge. The agent also knows things you don't (what they actually did all week). That information gap is what makes the problem hard.
A salesperson paid on bookings will close deals that book well and serve the customer badly. A contractor paid by the hour has no reason to finish quickly. A fund manager paid on this year's returns will take risks that pay off this year and blow up later, after the bonus clears. None of these people are villains. They are agents responding to the incentives a principal designed.
Principal–agent problem: when someone acting on your behalf has incentives that diverge from your goals (plus more information than you), they will optimize their incentives, not your goals. Aligned incentives are the fix — structuring rewards so the agent's self-interested choice is also the choice you want. Equity for early employees is the canonical example: their payoff is the company's payoff.
This connects directly to Lesson 4.8 (Power & Politics). Politics is, in large part, what happens when incentives are misaligned and information is hidden: people compete over the proxies that get them promoted rather than the outcomes the organization needs. Fix the incentive structure and a surprising amount of “politics” evaporates, because the self-interested move and the right move stop being different things.
FRAMEWORKDesigning incentives that survive contact with people
If gaming is the disease, here are the design principles that build resistance. Run any comp plan, OKR, or bonus scheme through these before you ship it:
Reward the outcome, not the activity
Pay for the thing you actually want, as far downstream as you can measure it. “Revenue retained” beats “calls made.” The closer the metric is to real value, the less room to game it.
Make it hard to game
Ask the killer question before launch: “If someone wanted to hit this number without doing the real work, how would they?” If there's an obvious cheat, assume it will be used.
Use multiple, balancing metrics
A single metric is always gameable. Pair it with a counter-metric: speed and quality, growth and retention, volume and satisfaction. Gaming one usually damages the other, which exposes it.
Balance individual and team
Pure individual incentives breed hoarding and internal competition; pure team incentives breed free-riding. Most healthy systems blend both so people help each other and still feel their own contribution.
Expect to revise it
Every incentive scheme decays as people learn to play it. Treat it as a living system, watch for the gaming, and adjust — don't set it and forget it.
When you set compensation, grant equity, or write OKRs, you are designing the behavior of your whole company for the next year. Equity aligns the long game (everyone shares the upside). Commission aligns the short game (and needs the most gaming-defence). OKRs that are tied to bonuses quietly become Goodhart targets — which is exactly why many strong teams keep OKRs as honest stretch goals and decouple them from pay.
WEAKNESSThe limits of bribery: when extrinsic incentives backfire
Here is the honest counter-weight to everything above: you cannot incentivize your way to everything, and trying can make things worse. Extrinsic incentives — money, prizes, points — have a documented failure mode called the overjustification effect. When you pay people to do something they already found intrinsically meaningful, the activity can re-frame in their mind from “something I do because I care” to “something I do for the reward.” Remove or shrink the reward and motivation can fall below where it started. You bought the behavior and, in doing so, you crowded out the reason they did it for free.
This is the dark mirror of Lesson 2.4 (Motivation): autonomy, mastery, and purpose drive the work that's hard to specify — creativity, judgment, care, citizenship. Those are precisely the things you most want and least can measure. Bolt a crude extrinsic metric onto them and you risk corroding the intrinsic engine that produced them in the first place. The research here (Deci, Ryan, and others on self-determination) is genuinely contested in its limits, but the practical lesson is robust: extrinsic incentives are powerful for clear, measurable tasks and risky for meaningful, hard-to-specify ones.
You can't write a bonus for “tells the truth in meetings,” “helps a teammate with no upside,” or “raises the unpopular concern.” Those run on culture and intrinsic motivation — and trying to incentivize them directly usually produces a hollow, gamed imitation. Some of the most important behaviors must be selected for, modeled, and protected — not paid for.
WEAKNESSWhere incentive thinking goes wrong
The default failure. You measure what's easy, reward the number, and people optimize the number instead of the goal (Goodhart's law in action). Before launching any metric, ask how you'd cheat it — then assume someone will.
Founders love to believe their team is mission-driven enough to override a bad comp plan. They aren't — not for long, and not under pressure. When the mission and the incentive disagree, bet on the incentive. If you want mission behavior, make sure the incentives don't punish it.
Over-engineered bonus schemes and points systems can crush the intrinsic motivation (the overjustification effect) that produced the good work in the first place. More metrics is not more control. Sometimes the right move is fewer incentives and stronger culture.
KEY INSIGHTWhere this connects in your course
- Motivation (2.4) — incentives are the lever; motivation is the engine. Extrinsic rewards interact with (and can corrode) intrinsic drive.
- Systems thinking (1.6) — incentives are an organization's strongest feedback loop, and the source of most unintended consequences.
- Power & politics (4.8) — much of what looks like politics is misaligned incentives plus hidden information. Align the incentives and politics shrinks.
- Accountability (3.9) — what incentives can't reach, a culture of ownership must. The behaviors you can't measure or pay for are exactly what culture is for.
Incentives sit at the heart of strategy because strategy is ultimately executed by people, and people execute what they're rewarded for. Design the incentives carelessly and your beautiful strategy quietly turns into whatever the comp plan actually pays for. Munger's line is not a clever aphorism — it's an operating instruction.
GLOSSARYKey terms introduced
- Incentive
- Anything — monetary or not — that makes a behavior more or less attractive to the chooser. Includes status, recognition, autonomy, security, promotion, and avoiding blame.
- Perverse incentive
- A reward structure that produces the opposite of the intended outcome by rewarding a gameable proxy.
- Cobra effect
- The folk name for a perverse incentive: a bounty meant to reduce cobras led to cobra breeding. Reward the proxy, get the gaming.
- Principal–agent problem
- When someone acting on your behalf has incentives that diverge from your goals and more information than you, so they optimize their incentives rather than your goals.
- Goodhart's law
- “When a measure becomes a target, it ceases to be a good measure.” Rewarding a metric decouples it from the work it once tracked.
- Aligned incentives
- Reward structures where the agent's self-interested choice is also the outcome the principal wants (e.g. equity for early employees).
- Gaming
- Hitting the rewarded metric without doing the real work it was meant to represent.
- Overjustification effect
- When an extrinsic reward replaces intrinsic motivation, so behavior drops once the reward is removed.
ACTIVE RECALLCheck your understanding
Answer in your own words first, then reveal.
1Why do people often follow incentives over instructions, mission, or logic?›
Because incentives shape behavior continuously while words shape it once, and following your incentives is rational, not cynical. When the mission and the reward disagree, the reward wins under pressure — so behavior is best predicted by what actually gets rewarded, not by what's stated.
2Explain the cobra effect and Goodhart's law in one breath.›
You can rarely measure the real goal, so you reward a measurable proxy. People then optimize the proxy instead of the goal — bounties breed cobras, line-count breeds bloat. Goodhart's law states the general case: when a measure becomes a target, it ceases to be a good measure.
3What is the principal–agent problem, and what does it mean to “align” incentives?›
An agent acting on a principal's behalf has their own incentives (and more information), so they optimize those, not the principal's goals. Aligning incentives means structuring rewards so the agent's self-interested choice is the principal's desired outcome — e.g. giving early employees equity so their payoff equals the company's.
4When should you NOT reach for an extrinsic incentive?›
For meaningful, hard-to-specify work driven by intrinsic motivation — creativity, judgment, honesty, helping teammates. A crude extrinsic reward can trigger the overjustification effect, crowding out the very motivation that produced the good work. Those behaviors belong to culture, not the comp plan.