Between 2011 and 2016, Wells Fargo employees opened as many as 3.5 million bank and credit card accounts that customers never requested, a scandal that cost the bank $185 million in regulatory fines and far more in trust. The cause was not a handful of dishonest tellers. It was a number. Branch staff were told to sell at least eight products per household, a goal leadership had boiled down to the slogan “Eight is Great.” That target was supposed to measure something real, namely how well the bank served each customer. The moment it became the thing people were paid and fired on, it stopped measuring service and started measuring how good employees were at opening accounts nobody wanted.

That is Goodhart’s Law in action, and once you have seen it clearly you start spotting it on your own team.

The law in one sentence, from two people who said it well

In 1975 a Bank of England economist named Charles Goodhart watched monetary indicators misbehave the instant the government tried to steer by them, and wrote the dry original: “any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.” Two decades later the anthropologist Marilyn Strathern sharpened it into the version everyone quotes: when a measure becomes a target, it ceases to be a good measure.

An American social scientist, Donald Campbell, reached the same place from inside schools and social programs and left us Campbell’s Law: the more any quantitative indicator is used for decision making, the more it gets corrupted, and the more it distorts the very process it was meant to track. Campbell added the part managers most need to hear. The higher the stakes riding on a number, the faster it rots.

The reason sits underneath both laws. Every metric is a proxy, a thin stand-in for a rich thing you actually care about. When nothing much rides on the proxy, people chase the real thing and the number follows along honestly. The instant the number carries real weight, rational people optimize the number directly, by the shortest path available, and that path almost never runs through the outcome you wanted.

It is not a banking problem. It is a measurement problem.

In 2004 England’s National Health Service set a target: 95 percent of patients arriving at accident and emergency should be admitted, discharged, or transferred within four hours. The intent was humane, since long waits in emergency rooms genuinely harm people. What followed was a catalog of gaming documented in the medical literature. Patients were admitted at the three-hour-58-minute mark whether or not admission actually helped them, ambulances sometimes idled outside so the clock would not technically start, and admissions spiked right before the four-hour line. One widely cited paper summed it up as “hitting the target but missing the point.” Care improved far less than the number did, because staff were now managing a clock instead of a patient.

Swap the setting and the pattern is identical. A bank, a hospital, a sales floor, a factory, even the old colonial bounty on cobras that quietly spawned a cobra-breeding industry (a story economists love, though its details are disputed). The mechanism never changes. Point a strong enough incentive at any single figure and people will move that figure, cheaply and creatively, and the thing it was supposed to represent gets left behind.

I watched it happen on my own floor

Running IT operations at a large telecom, I once put a target on average ticket handle time to push the service desk to be faster. Within a quarter the dashboard looked excellent; average handle time had dropped and the trend line was green. Then the reopen rate started climbing, and so did repeat calls from the same customers about the same problems. The analysts had learned the system exactly as designed. The dashboard watched time-to-close and nothing else, so the fastest way to a good number was to close tickets quickly and open a fresh one when the customer called back, rather than to fix the problem the first time.

We had measured speed and gotten the appearance of speed. The work underneath got worse, not better. I had not hired careless people that quarter; I had pointed careful people at the wrong number and they did what the number asked. That is the humbling part of Goodhart’s Law. It does not require anyone to be dishonest. It just requires them to be responsive, which is the quality you hired them for.

What a manager actually does about it

You cannot stop measuring. Running a team blind is worse than running it on imperfect numbers. What you can stop doing is asking a single figure to carry more weight than a proxy can bear. A few habits hold up under pressure.

Pair every target with its guardrail. If you reward speed, watch quality in the same breath. Handle time sits next to reopen rate. Sales volume sits next to complaints and chargebacks. Shipped features sit next to defects and rollbacks. A number is reasonably safe to chase only when a second number would catch the cheat, because the easy way to game the first one shows up as a spike in the second.

Measure outcomes, not activity, wherever you can. Activity metrics (calls made, tickets touched, story points burned) are the easiest to inflate without doing anything real. Outcome metrics (the problem stayed fixed, the customer renewed, the feature actually shipped and worked) are harder to fake because faking them tends to require doing the real thing. When you set goals specific enough to drive results, anchor them to what you want to be true, not to the motion that usually precedes it.

Keep your highest-stakes numbers off individual scorecards. This is Campbell’s warning made operational. Use metrics to start conversations, not to trigger automatic rewards and punishments. The second a bonus or someone’s job hangs on one figure, you have told that person precisely which number to defend at the expense of everything around it. Wells Fargo did not have a people problem. It had a compensation system wired directly to a single proxy.

Ask your team how they would game it, out loud, before you launch it. The people who will live under a metric already know its loopholes. Ask them plainly: if you wanted to hit this without doing the real work, how would you? They will usually tell you, often with a grin, and every answer is a hole you can close before it costs you a quarter. It is one of the cheapest things you can do to protect the output you actually care about.

A target is a useful servant and a terrible master. The figure on your dashboard was always a stand-in for something you cannot fit on a dashboard: a customer genuinely helped, a problem actually solved, a product people want to keep paying for. Goodhart’s Law is the standing reminder that the stand-in and the real thing stay aligned only as long as nobody is desperate to move the stand-in. Your job is not to hunt for the one perfect metric, because there isn’t one. Your job is to keep the numbers honest by never asking them to do more than a number can.