Goodhart’s Law and why chasing metrics often ruins the actual work

Tell a crew they’re getting paid by the number of nails they drive, and you’ll end up with a house held together by a million tiny tacks. That’s Goodhart’s Law: the second a metric becomes the target, it stops being a good way to measure anything.

People aren't stupid; they’re just efficient. If you judge a call center by how fast they hang up, don't be shocked when they start cutting off customers mid-sentence. The goal was "efficiency," but the metric just rewarded being rude.

We think we’re tracking progress, but we’re usually just teaching people how to cheat the scoreboard without actually doing the work.

Wait, so is it actually possible to measure success without ruining it?

It’s tricky, like trying to judge a BBQ solely by the temperature of the grill. You might have the heat dialed in, but if the snags are burnt to a crisp, you’ve still failed the mission.

The secret is to never lean on a single number. You need a "basket" of metrics that pull in opposite directions. If you’re tracking how fast the crew works, you’d better be tracking how many mistakes they make, too.

It’s about the "vibe check" versus the spreadsheet. When you use multiple data points that conflict, it’s much harder for someone to cheat the system without the whole house of cards falling over.

Surely tracking all those extra numbers just creates a massive mountain of paperwork?

It’s not about counting every grain of sawdust. If you try to measure everything, you’ll spend all day with a clipboard and zero time actually building the deck.

The trick is finding the "tension." Think of it like a tug-of-war. If you only measure speed, the rope flies off and quality hits the dirt. You just need one or two "counter-weights" to keep the rope centered.

It’s about a few smart guardrails, not a mountain of paperwork. You're just making sure the "fast" guy isn't also the "dangerous" guy.

But how do you actually spot two metrics that'll balance each other out?

You’re looking for natural enemies, like a cat and a dog. If your metrics are too friendly, they’ll just team up to hide the truth from you.

If you’re measuring "how many holes we dug," the counter-weight isn't "how much dirt we moved"—those are teammates. It’s "how many underground pipes we accidentally burst."

The goal is friction. When one number goes up, the other should get harder to keep steady. If they both rise easily, you haven't found a balance—you've just found two ways to measure the same shortcut.

Doesn't creating that much friction just paralyze the whole operation?

Paralysis only happens if you crank the tension too tight. You’re not trying to lock the brakes; you just want enough resistance so nobody flies off a cliff.

Think of it like riding a bike downhill. Slamming the brakes stops you completely, but a light squeeze keeps you moving safely.

That friction is just there to force a brain-check. It tells the crew when to slow down and pay attention, instead of blindly rushing to hit a dumb target.