Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.

Charles Goodhart

(Or put simply: Any measure stops being useful if you make it into a target.)

The action: If you turn a metric into a goal (and reward people for reaching that goal), beware of non-intended side effects. Stay close to the action, and help people improve the underlying system.

The long form: Cobras were becoming a problem in colonial Delhi. They were everywhere, causing harm to humans and animals, so Indian and British authorities had to step in and cull the snake population.

Killing snakes seemed like a good job to outsource. After all, the results – a dead snake – would be easy to count and to reward, exactly the type of work often recommended for “pay-for-performance”-programs. So the government introduced a system where you could bring dead cobras to the local police office, and receive a bounty for each one.

This is management by objective. “I don’t care how you do it, as long as you get the job done”.

Now, this form of intention-driven management is not bad by itself. It is often seen as the antidote to micromanagement – a form of leadership in which you rob people of their creativity and pride in workmanship. However, you should be aware that there are only three ways to reach a goal such as this:

1. Improve the system
2. Distort the system
3. Distort the data

In the cobra-case, the authorities were clearly hoping the cash-for-snakes-scheme would lead to a system improvement – fewer cobras. However, it turned out a few enterprising individuals realized what a gold mine this could be, and started breeding cobras for cash, distorting the system. While the city was paying handsomely to reduce the number of snakes, the cobra population was actually increasing!

The cobra anecdote is often cited as the classic example of the perverse incentive effect – an outcome very different to what the designer of the system intended. And this happens everywhere in our work lives too. Once a reduction in the number of snakes is no longer a measure, but a goal, Goodhart’s Law states that it introduces the risk of gaming the system and thus stops being useful.

Think of national tests for schools. Their original intention is good: to find out how our schools are doing and be able to compare them to other schools. However, once improving the test results becomes a goal in itself, teachers spend more time preparing students for tests instead of teaching the regular curriculum. This can even lead to a situation where the test results show which schools have sacrificed the most teaching time for test preparation, rather than which schools are better at teaching in general.

The trouble with external incentives is not that they don’t work, but that they often work too well, crowding out other worthwhile goals.

For some, this has led to an allergy to targets and goal-setting of any kind. But perhaps a better way is to not “fire and forget” an objective, but to stay close to the people coming up with solutions and improving the system. Instead of saying “I don’t care how you fix this,” you create a rhythm of improvement, set aside enough time and also coach the work. In this way you make it easier to improve the system than to distort it.