Turns out, the easiest way to multiply a problem is to start paying people to solve it.
British colonial rulers in India faced a major problem in Delhi. Venomous cobras were everywhere, putting residents at risk. To fix the issue quickly, the government introduced a straightforward economic solution. They offered a cash bounty for every dead cobra brought to them. At first, the policy worked brilliantly. Locals hunted the snakes, turned them in, and the wild cobra population dropped.
But as wild cobras became harder to find, people realized a flaw in the system. The British weren't paying for hunting cobras. They were paying for dead ones. To keep the cash flowing, enterprising locals set up breeding programs in their homes. They raised thousands of venomous snakes strictly to kill them and collect the reward.
Eventually, administrators noticed something strange. The number of dead snakes coming in stayed massive, even though the wild population was clearly depleted. Realizing they were funding a thriving snake breeding industry, the government canceled the bounty program overnight.
The breeders were suddenly stuck with thousands of worthless, dangerous snakes. Unwilling to pay to feed them, they released the cobras into the wild. The city ended up with a much worse snake problem than when the whole thing started.
British colonial rulers in India faced a major problem in Delhi. Venomous cobras were everywhere, putting residents at risk. To fix the issue quickly, the government introduced a straightforward economic solution. They offered a cash bounty for every dead cobra brought to them. At first, the policy worked brilliantly. Locals hunted the snakes, turned them in, and the wild cobra population dropped.
But as wild cobras became harder to find, people realized a flaw in the system. The British weren't paying for hunting cobras. They were paying for dead ones. To keep the cash flowing, enterprising locals set up breeding programs in their homes. They raised thousands of venomous snakes strictly to kill them and collect the reward.
Eventually, administrators noticed something strange. The number of dead snakes coming in stayed massive, even though the wild population was clearly depleted. Realizing they were funding a thriving snake breeding industry, the government canceled the bounty program overnight.
The breeders were suddenly stuck with thousands of worthless, dangerous snakes. Unwilling to pay to feed them, they released the cobras into the wild. The city ended up with a much worse snake problem than when the whole thing started.
The Metric Trap
Known as the Cobra Effect, it perfectly illustrates a perverse incentive. When you tie a reward to a specific metric, people will optimize for that metric rather than the actual goal.
The British measured success by dead snakes, assuming it meant a safer city. By paying for the output instead of the outcome, they accidentally incentivized the exact opposite of what they wanted. Systems always prioritize what gets measured and rewarded, no matter how disconnected it gets from reality.
The British measured success by dead snakes, assuming it meant a safer city. By paying for the output instead of the outcome, they accidentally incentivized the exact opposite of what they wanted. Systems always prioritize what gets measured and rewarded, no matter how disconnected it gets from reality.
Modern Snake Farms
This same dynamic plays out in business and software development all the time.
Take software bug bounties. If a company pays developers a cash bonus for every bug they find and fix, they often get a sudden surge in reported bugs. Why? Because developers quickly realize they control both the creation and the fix. Some start writing sloppy code on purpose, slipping in trivial errors they can easily log and resolve to collect extra cash. The software ends up more unstable, while the company bleeds money fixing manufactured problems.
Or look at customer support metrics. If management wants to reduce customer wait times, they track average handle time and penalize agents who stay on the phone too long. Agents figure out the fastest way to close a ticket isn't to solve the problem. They start transferring angry customers, giving rushed answers, or hanging up before they even finish a sentence. Call times drop, but customer satisfaction tanks and repeat callers skyrocket.
When you design an incentive structure, you have to look past the surface level metric. Ask yourself what shortcuts people can take to game the system. If there's an easier path to the reward that doesn't actually solve the underlying problem, people will take it every single time. Stop rewarding the output. Focus on the outcome, and build systems that make it impossible to profit from the problem.
Take software bug bounties. If a company pays developers a cash bonus for every bug they find and fix, they often get a sudden surge in reported bugs. Why? Because developers quickly realize they control both the creation and the fix. Some start writing sloppy code on purpose, slipping in trivial errors they can easily log and resolve to collect extra cash. The software ends up more unstable, while the company bleeds money fixing manufactured problems.
Or look at customer support metrics. If management wants to reduce customer wait times, they track average handle time and penalize agents who stay on the phone too long. Agents figure out the fastest way to close a ticket isn't to solve the problem. They start transferring angry customers, giving rushed answers, or hanging up before they even finish a sentence. Call times drop, but customer satisfaction tanks and repeat callers skyrocket.
When you design an incentive structure, you have to look past the surface level metric. Ask yourself what shortcuts people can take to game the system. If there's an easier path to the reward that doesn't actually solve the underlying problem, people will take it every single time. Stop rewarding the output. Focus on the outcome, and build systems that make it impossible to profit from the problem.