Why a measure stops being useful once it's a target

HostI was reading this wild story about a nail factory back in the Soviet days. If the bosses told the workers they needed to make a huge number of nails, the workers would just churn out millions of tiny, little pins that were too small to hold anything together. But if the bosses switched the rule and said they were measuring success by total weight, the workers would make these massive, heavy spikes that were basically chunks of iron. They looked like nails, but you couldn't actually use them for anything. It feels like this strange trap where the moment you try to measure how well a job is going, you end up ruining the very thing you wanted to achieve. Is this just a quirk of old factories, or is there something deeper going on?

GuestIt's much deeper than that. It happens in almost every big system, whether it's a hospital, a school, or a huge global company. The problem is that we rarely measure the thing we actually care about. Think about it. We care about "patient health" or "good schools," but those ideas are too messy and abstract to just count up on a sheet. So, we find a stand-in, which is often called a proxy. It's a bit of data that we can actually see and track. For example, a hospital might care about how healthy people are, but they measure how many people have to come back within a month of leaving. That's a proxy. The big mistake we make is thinking that the number is the same thing as the goal. There's this idea called Goodhart’s Law that says a measure is only useful as long as you're just using it to watch the system. The second you turn that measure into a goal you have to hit, it stops being a window into the truth and starts being a mask that hides what's really happening.

HostBut wait, if you want people to stay healthy, isn't it good to track if they're coming back to the hospital? Why does the link between that number and the truth just break because you started looking at it?

GuestBecause the relationship between the number and reality only holds up when things are running naturally. When you turn that number into a target, you change how everyone in the system behaves. Humans are like heat-seeking missiles for rewards. We're incredibly good at finding the easiest path to hit a goal, even if it means doing something silly. Take the city of Hanoi in 1902. The officials there wanted to get rid of rats, so they offered a small cash prize for every rat tail someone brought in. They thought, great, people will catch all the rats. But people weren't just catching wild rats. They started breeding rats in their backyards just to cut off the tails for the money. They would even catch a rat, chop off the tail, and let it go back into the sewers so it could've more babies and make them more money. The city ended up paying for thousands of tails while the rat problem actually got worse. This is what we call gaming the system. It's not even about people being bad or dishonest. It's just a rational reaction to the pressure. People will always move toward the specific number they're being judged by, even if it hurts the original plan.

HostSo it's basically a lose-lose. If you set a goal, people find a way to cheat it, and if you don't set a goal, you have no idea what's happening. It sounds like you can't win.

GuestIt feels that way because of how much we rely on these numbers for big life-changing decisions. This is where we run into Campbell’s Law. It says that the more a high-stakes decision—like a school getting its funding or a person keeping their job—depends on one single number, the more that number will be corrupted. We see this all the time in schools with "teaching to the test." A test is supposed to be a tiny sample to see if a kid is learning. But when the test score is the only thing that matters, that tiny sample becomes the entire school year. Teachers stop teaching the big, rich ideas and start teaching the narrow tricks to pass that one specific exam. This leads to something called goal displacement. You hit the target perfectly. The scores go up. But the actual mission of the school, which is to help kids think and learn, gets totally ignored. People start caring more about the appearance of success than the actual substance of their work.

HostIf these numbers are so easy to mess with, should we just stop using them to decide who gets funded or who gets a bonus?

GuestNot necessarily, but we have to stop letting them stand alone. The best way to fix this is to use many different, often clashing numbers to look at the same goal. Experts call this triangulation. Think about a call center. If you only track how fast people hang up the phone, your staff will just rush every caller to keep their times low. But if you also track if the customer is happy and if their problem actually got fixed on the first call, you create a kind of healthy tension. You can't cheat the speed goal without hurting the satisfaction score. By looking at the goal from many angles at once, you make it much harder to game the system. Most importantly, we have to start using data as a way to learn and start a conversation, rather than just using it as a hammer to force people to change. Data should be read-only as much as possible. It's a way to say, hey, what's actually going on here, rather than a rigid rule that replaces human judgment.

GuestWhen you make one single number the only thing that matters, you lose sight of why you started the work in the first place.

HostIt's just like those heavy iron spikes from the factory — we end up getting exactly what we asked for, only to realize it's not what we actually needed.