Pay for Rat Tails, Get Rat Farms

The full Hanoi story

Hanoi's rat hunt is documented in the colonial archives, found by the historian Michael Vann. The French sewers, more than nine miles of pipe by 1902, were a symbol of modern progress. Underground, nothing hunted the rats, and the pipes led straight up into the best villas in town.

Plague was moving through Asia, and French researchers had tied it to rats and their fleas. So in 1902 the city first paid Vietnamese hunters for every rat they killed in the sewers. The archived tallies climb from 7,985 rats in the last week of April to 20,112 on a single day in June. It still wasn't enough, so the city opened the hunt to everyone at one cent per rat, tail only.

The bounty was scrapped and plague came anyway: an outbreak that began in 1906 killed at least 263 people in Hanoi, most of them Vietnamese. Even the numbers were soft. Vann says the daily rat counts, like most colonial statistics, were "really a best guess."

Vann warns against reading Hanoi as a case against intervention. His lesson is to watch for programs built where power is so lopsided that evidence gets ignored. In a business, that's the bonus plan your front line could have gamed in five minutes, if anyone had asked them first.

This is the Data loop at work: change what you measure, and People, Process and Technology move with it, whether you planned for it or not.

Why do smart people build bad incentives?

Three old ideas explain almost every case above.

When a measure becomes a target

In 1975 the British economist Charles Goodhart observed that a statistic stops behaving once policy starts leaning on it. The anthropologist Marilyn Strathern later put it in the words most people know: "When a measure becomes a target, it ceases to be a good measure." This is Goodhart's law.

Rewarding A while hoping for B

That's the title of a 1975 paper by the management scholar Steven Kerr, still widely reprinted. People work out what actually gets rewarded and do that, often to the exclusion of everything else. His first two suspects are our love of simple, countable standards and our habit of rewarding what's easy to see, which leaves things like teamwork and creativity unrewarded.

Rule beating and the wrong goal

The systems scientist Donella Meadows named both as traps in Thinking in Systems. Rule beating is behavior that looks like compliance but defeats the purpose of the rule. Seeking the wrong goal is a system faithfully chasing an indicator that doesn't reflect what anyone actually wants. Her way out: treat rule beating as feedback, and choose indicators that track the real goal, not the effort spent on it.

Measure like a barbell, or breed cowards

Most scorecards punish every miss. That sounds like accountability. In practice it teaches people to set goals they know they'll hit and to skip any experiment that might fail. Kerr saw this too: companies say they want challenging stretch objectives, then reward making the numbers.

Investors have a better shape for risk, which Nassim Nicholas Taleb made famous as the barbell: most of the money in very safe bets, a small slice in bold ones, almost nothing in the middle. A loss on the bold side is survivable. A win can change everything.

I think people should be measured the same way. Hold the core of the job to steady, dependable numbers. Then protect a small slice of time for bets that are allowed to fail, and judge that slice on a different yardstick, such as what the team learned and how fast. Measure everything with the same ruler and you breed cowards.

More on AI reward hacking

In OpenAI's 2025 report, the models often said what they were doing plainly in their reasoning: "Let's hack." One rigged a verification step so it always returned true. Penalizing the visible thoughts didn't stop the behavior; it hid it.

That should sound familiar. Wells Fargo employees changed customers' contact details so they wouldn't learn about accounts opened in their names. Put pressure on the visible sign and the behavior moves out of sight.

The agency story, in full

Department heads were told to shrink the bench and make everyone billable. Most teams were already at or over capacity, because projects were rarely estimated well, and people were burning the candle at both ends. Meanwhile the account and business development team was told to do whatever it took to win new clients and grow existing ones.

Those two groups had always worked well together. Then revenue, and chasing every client, became the top priority, ahead of taking care of the client's own customers. So everyone watched their own back instead of protecting their teams and the people the work was really for. It became about how much each person could get out of it, and how hard the team could be squeezed.

FAQ

What is a perverse incentive?

A perverse incentive is a reward that produces the opposite of what it was meant to, because people can earn it by producing a sign of the result instead of the result. Hanoi's rat tail bounty, which led to rat farms, is a documented example.

What is the cobra effect?

The cobra effect is a popular name for a perverse incentive, taken from a story about snake bounties in British Delhi. The Hanoi rat hunt of 1902 is a documented case of the same pattern.

What is Goodhart's law?

Goodhart's law says that when a measure becomes a target, it ceases to be a good measure. Charles Goodhart made the observation in 1975; Marilyn Strathern gave it that wording.

How do you design incentives that don't backfire?

Start from the real outcome, ask the people being measured how they could game the number, pair every number with a counterweight, check whether the reward makes anyone want the problem to last, and review the results on a set date so gaming becomes feedback on the rule.