The story behind the name
- Step 1Authorities offer a reward for every venomous snake killed: there are too many snakes in the city.
- Step 2Residents start breeding snakes on purpose so they can turn them in for the reward.
- Step 3The program is canceled: costs are rising, but there are no fewer snakes in the streets.
- Step 4The bred snakes are released, and there are more of them than there were at the start.
Historians treat this story with caution: there is little documentary evidence, and it sounds more like a parable. But a similar case involving rewards for rat tails in colonial Vietnam is much better documented.
People turned in tails rather than carcasses, and soon they began spotting rats without tails in the city: trappers were releasing them so they could keep breeding. Some enterprising people were even breeding rats outside the city.
How to spot this trap in advance
- The metric can be produced: the result can be created not only by solving the problem, but also by growing it.
- Only part of the whole is measured: tails instead of rats, calls instead of resolved requests, lines of code instead of a working program.
- The reward is large relative to the effort: the more profitable the metric, the faster someone finds a workaround.
- There is no on-site check: no one looks to see whether the original situation has changed.
This is often summed up as a rule: once a measure becomes a target, it stops being a good measure. Not because people are dishonest, but because any measure turned into a reward starts taking on a life of its own.
You can easily recognize modern examples. An ambulance service judged by arrival time starts sending the nearest vehicle instead of the most suitable one; a hospital judged by mortality becomes more cautious about taking critically ill patients.
The answer is not to eliminate metrics, but to use several and check them. Use different measures, carry out random on-site checks, and review them regularly - otherwise the system optimizes exactly what you measure, in every sense.
What is the essence of the cobra effect?
What signals that a metric is dangerous?
Sources
- The term "cobra effect" was popularized by economist Horst Siebert
- documentary evidence for the snake-reward story itself is limited, while the case involving rewards for rat tails in colonial Hanoi is described in historical research. The phenomenon is linked to Goodhart's law: once a measure becomes a target, it ceases to be a reliable measure
Next in the “Great Mistakes” series
2 articles in full on the site, 3 more in the app
