The moment you pay for the number, the number stops telling you the truth.
A school district wants better teachers, and it has a reasonable way to tell them apart: students of good teachers tend to score higher on tests. So the district starts paying for scores. Within a few years scores are up across the board, and the students know less. What went wrong? The measure was accurate before anyone was rewarded for it. Nothing about the test changed.
The usual answer is that the district picked a bad metric and should find a better one. That answer is wrong, and the reason is the point of Goodhart’s Law: When a measure becomes a target, it ceases to be a good measure.
Why the number comes loose
Test scores tracked good teaching only because nobody was pushing on them. A good teacher raised scores as a side effect of helping students understand, so the number and the goal moved together. Once the number carries a reward, every teacher looks for the cheapest way to move it, and the cheapest way is almost never the goal. Drilling the test format is cheaper than building understanding. So the number keeps rising while the thing it used to indicate drifts away.
The mistake underneath is reification: treating the proxy as if it were the objective. A metric is a window onto something too complicated to watch directly. Reward people for the window and they will paint on the glass.
The damage is also hard to see, because the one instrument you were using to watch the school now reports success. To argue that the school is getting worse, you need either a different number, which will be gamed the moment it carries weight, or direct observation of classrooms, which is exactly what the metric was adopted to avoid.
That is why a cleverer metric doesn’t escape the law: the problem is not in the quality of the metric but in the act of targeting it. A sophisticated measure gets gamed more cleverly, not less. The durable defense is never forgetting that the metric is a proxy, not the thing itself, and asking, before optimizing a number, whether it should be a target at all.
What a metric admits
Why did the district reach for test scores in the first place? Because it did not trust principals or teachers to judge teaching, and could not afford to watch every classroom itself. Most metrics begin this way, as a substitute for trust or attention. The deeper you rely on a metric, the more you are admitting that you have substituted measurement for understanding.
That substitution is sometimes necessary. Nobody can understand every system they manage, and a number often beats a hunch: it can expose favoritism, catch a failing school, or show that a cherished program does nothing. The honest response is humility about the numbers rather than abandoning them. Read them, and then go and look at what they are standing in for, especially when they look good.
The proxies you set for yourself
The same law runs inside a single life, where nobody is gaming the measure on purpose. Years ago you chose a salary, a title, or a weight as a stand-in for something you actually wanted, such as security, respect, or health. The number was a reasonable indicator then. Then it became the thing you checked, and checking it became the thing you optimized. That is how bad decisions arise from long unexamined proxy optimizations.
The district at least has a parent who notices that the children can’t read. In your own life, the person who might notice that the number has come loose from what it stood for is usually you. “What is it good for?” is the question that notices. Asked of the number you watch most closely, it may give an uncomfortable answer.
References:
- Venkatesh Rao, Goodhart’s Law and Why Measurement Is Hard, Ribbonfarm