Goodhart’s Law, the cobra bounty, and what happens when a measure becomes a target
Somewhere in your house, possibly on your wrist right now, there may be a device counting your steps. Ten thousand a day is the number, and it is a perfectly reasonable stand-in for “move your body a bit more”.
Right up until the evening you find yourself pacing the kitchen at half past eleven, not because you want to walk, but because you are 800 steps short and the ring hasn’t closed. The number will go up. Whether you are any healthier for the lap of the fridge is a different question entirely. The step count was only ever standing in for something else, and the moment hitting it became the goal, the two quietly parted company.
Last time in this series we looked at the numbers that mislead you by leaving things out – the bombers that never came home. This is almost the opposite problem: what happens when a number is so visible, so watched and so important that people start aiming straight at it.
It has a name.
When the measure becomes the target
Goodhart’s Law is usually stated like this: when a measure becomes a target, it ceases to be a good measure.
The phrasing belongs to the anthropologist Marilyn Strathern, who was tidying up a drier observation by the economist Charles Goodhart in the 1970s. Goodhart was writing about monetary policy, but the idea turned out to describe almost everything. Once you take a number that was quietly describing the world and start using it to steer the world, people respond to the number itself, and it stops describing much of anything.
The trap is that this rarely involves anyone doing something obviously wrong. It just involves people being sensible about the target in front of them.
The cobras of Delhi
The story usually told to illustrate this comes from colonial-era Delhi.
The administration, so the tale goes, was worried about the number of venomous cobras in the city and hit upon an elegant solution: pay a bounty for every dead cobra brought in. Citizens would be incentivised to kill snakes, the population would fall, everyone wins. And at first it worked.
Then some enterprising residents noticed something. If a dead cobra was worth money, a live cobra was a small business opportunity. They began breeding cobras specifically to kill them and claim the bounty. The authorities eventually caught on and scrapped the scheme, at which point the breeders, now sitting on stock of no value, released their cobras. The city ended up with more snakes than it had started with.
The bounty was a superb measure of “dead cobras handed in”. It was a hopeless measure of “fewer cobras in Delhi”, the thing anyone actually cared about, the moment there was money in moving it. It has since lent its name to the whole phenomenon: the cobra effect.
Whether every detail of that story is precisely documented is debated. It survives because the pattern underneath it is real and instantly recognisable.
Why it happens
Underneath Goodhart’s Law is a simple mechanism. Almost every number we track is a proxy. It stands in for something we care about but can’t measure directly. Satisfaction stands in for a good experience. A pass rate stands in for real learning. Steps stand in for health.
A proxy works beautifully as long as nobody is leaning on it. The trouble starts when you attach stakes – money, reputation, a regulatory threshold, a league table position – to the proxy itself.
Because there are always two ways to hit a target. You can do the hard, genuine thing the number was standing in for. Or you can find the shortcut that moves the number without moving the thing. The second route is almost always faster, cheaper and less painful, and it does not feel like cheating from the inside. It feels like being responsive to your priorities. That is exactly why it is so hard to resist, and why the link between the measure and the meaning tends to snap under pressure.
Closer to home
Higher education runs on proxies, and most people working in it can feel this one in their bones.
Take a student satisfaction score. As a quiet description of how a course is landing, it is genuinely useful. Turn it into a target that feeds league tables and internal scrutiny, and a second set of options appears alongside “improve the course”: time the survey carefully, remind students how much they have enjoyed things just before they fill it in, nudge the framing. None of it is fraud. All of it moves the number without necessarily moving the experience the number was meant to capture.
Or take a continuation metric with a regulatory threshold attached. One way to improve the share of students who continue is to support the ones who are struggling. Another, quieter way is to be a little more cautious about who you admit in the first place, or a little more careful about how a withdrawal gets recorded. The number improves either way. Only one of those routes helps an actual student.
The point is not that institutions are gaming their figures. It is that the pressure to is real, structural and always present, and that clearing the harder road takes a deliberate, ongoing choice. The measure will not make that choice for you. If anything, it gently pushes the other way.
What to ask instead
You cannot escape targets, and you would not want to – measured well, they are how organisations know whether they are getting anywhere. But you can hold them at the right arm’s length. A few questions help:
What is this number actually a proxy for? Name the real thing out loud. If you can’t, that is worth knowing before you start chasing the number.
Could this metric improve without the real thing improving at all? If yes, you have found the gap someone, somewhere, will eventually walk through, usually without meaning any harm.
What behaviour does attaching stakes to this actually reward? Not what you hope it rewards. What it rewards.
And the big one: are we still using this number to watch reality, or have we started using it to drive it? Because those are different jobs, and a number is rarely good at both.
The point
Goodhart’s Law isn’t an argument against measuring things. It is a warning about what happens to a measurement when it stops being a mirror and becomes a lever. The number can go up, the target can be met, the report can look excellent, and the thing you actually cared about can be quietly going nowhere, or backwards.
So when a key figure has become a target that everyone is working hard to hit, it is worth asking what it was originally meant to stand for, and whether it still does. Because yes, the target might have been met. But what does it actually mean?