So you've picked your north star metric. You've announced it to the team. Everyone knows the number, everyone's aligned, feels good.

What happens the next day?

Honestly, probably nothing bad. First few weeks are fine. People are energized, focused, moving in the same direction. The number starts going up. Board meeting goes well.

But somewhere around month three or four, something subtle starts happening. The tactics that move the metric and the tactics that actually build a healthy business start to quietly drift apart. Not dramatically. Just a little. And then a little more.

There's a name for this. Goodhart's Law. A British economist named Charles Goodhart figured it out in 1975 studying monetary policy: when a measure becomes a target, it ceases to be a good measure. The moment you point your whole organization at a number, the number stops accurately representing the thing you were trying to measure. People, not out of malice usually, find the path of least resistance to hitting it.

Together With

Free Executive 1:1 Startup Growth Review

Guessing your way through early-stage growth is how startups often stall (or fail).

Get 20 minutes with senior marketing specialists who will:

Metrics get gamed in three distinct ways, and understanding which one you're dealing with changes everything.

The first is innocent gaming.

In 2018 Zuckerberg decided Facebook had a passive consumption problem. People were scrolling, watching videos, not actually connecting. So he changed the algorithm to chase "meaningful social interactions," comments, shares, reactions between real people. Genuinely well-intentioned.

Engagement went up 50% in the first year.

The product got worse.

The fastest path to comments and shares wasn't warmth or connection. It was outrage. Divisive posts, misinformation, stuff that made you angry enough to respond, these scored highest on every meaningful interaction the algorithm could measure. Internal research confirmed it was rewarding outrage. Facebook's own people knew.

Nobody cheated exactly. The metric just found the shortest path to itself.

You see this in startups constantly btw. Team picks DAU as the north star, someone notices push notifications move the number, so they send more, DAU climbs, six months later users are disabling notifications or just leaving. The team is genuinely confused. They were doing their jobs.

The second is pressure gaming.

Wells Fargo decided its key metric was the cross-sell ratio, basically how many financial products each customer held. The target was eight products per customer. Why eight? The CEO later told Congress it was because the number rhymes with great. That's the actual answer he gave.

So now you've got branch employees, most of them making around $12 an hour, being told to hit a number their CEO picked because of a rhyme. Miss quota and you get berated. Miss it again and you're on a plan. Keep missing and you're fired. Branch sales numbers were reported to management up to seven times a day.

So people started opening accounts without customers knowing. By the time regulators caught up, there were 3.5 million fraudulent accounts, 5,300 fired employees, and a $3 billion settlement. Wells Fargo increased its cross-sell ratio per household from 3.4 in 1999 to 6.11 in 2015. Nomura estimated that the ratio would have been reduced by just 0.1 products per household had the unauthorized accounts never been created. Basically nothing. The gaming didn't even work.

What the metric actually did was create enough pressure that ordinary people did things they never would have done otherwise, for almost no measurable gain. That's a different kind of failure than gaming the number successfully. It's worse, actually.

You're probably thinking that's a bank, different situation. But the same dynamic happens at 15 people. Board meeting coming up, deal is close, customer is clearly a bad fit, sales signs them anyway. ARR goes up, looks great. Twelve months later churn spikes and nobody connects it back to that one decision made under pressure.

The third is slow gaming.

This one is hardest to see because it looks like good execution for a long time.

The team gets genuinely good at moving the number. Playbooks get built around it, instincts sharpen. But slowly, like 18 months slowly, the tactics that move the metric start drifting from the tactics that build a healthy business. Engagement goes up because the product got better at triggering habit loops, not because users love it more. ARR grows but renewals are getting done with quiet discounts that don't show up in the headline. The metric is technically accurate. It just stopped meaning what it used to mean.

This version often doesn't feel like a problem until it's a big problem. Because the number keeps going up.

Every headline satisfies an opinion. Except ours.

Remember when the news was about what happened, not how to feel about it? 1440's Daily Digest is bringing that back. Every morning, they sift through 100+ sources to deliver a concise, unbiased briefing — no pundits, no paywalls, no politics. Just the facts, all in five minutes. For free.

Read the newsletter trusted by 4.5 million fact-seekers.

Why founders are specifically exposed

A few reasons that I don't think get enough airtime.

Speed is one. Small teams move fast and when something is clearly the number, people optimize without a lot of second-order thinking. In bigger organizations there's enough friction that someone usually asks whether the metric is moving for the right reasons. At a startup that pause feels like a luxury you can't afford.

The board dynamic is the other one and it's underrated. Every investor meeting, the number gets reported. The metric stops being internal truth and becomes external performance. The moment you're managing the number partly for optics, you've introduced a distortion that's really hard to undo. Boards aren't trying to create this pressure. The structure of the relationship just does it automatically.

What to actually do

Look — the answer is not to go back to tracking 80 things. That just creates a different problem where nothing moves and nobody owns anything.

A few things that actually help:

Pair your north star with one shadow metric that captures the downside. Optimizing DAU? Watch unsubscribe rate and notification opt-outs too. Optimizing ARR? Watch 12-month churn. The shadow metric doesn't go in the all-hands. It doesn't become a target. It lives on a private dashboard and its only job is to tell you when the main number is being gamed — whether anyone meant to or not.

Keep some signals private. When a number is widely known and tied to rewards, people optimize for it. That's not a character flaw, it's just how incentives work. Some signals stay honest longer when fewer people know about them. This isn't about hiding stuff from your team, it's about being deliberate about which information shapes behavior.

Be willing to retire the metric before the team fully games it. Metrics have a natural lifespan. The number that told you something true at 10 employees might not tell you the same thing at 50. Most founders make this switch reactively — when they notice things feel off. The better move is proactively, while it still feels useful, before the organizational muscle memory around gaming it runs too deep.

The thing about maps

Goodhart's Law doesn't mean metrics are bad or that you shouldn't pick a north star. It means the map is not the territory.

Your number is a proxy for something real — user love, business health, whether you've actually built something people need. It is not the thing itself. The moment you forget that, the metric starts working against you quietly and efficiently and usually while looking great on a dashboard.

The founders who handle this well aren't the ones who pick better metrics. They're the ones who stay close enough to the actual ground truth — talking to customers, watching real behavior, reading signals that haven't been formalized yet — that they notice when the number and the reality have quietly split apart.

By the time that happens the number will probably look fine.

That's the tell.

If this was useful, share it with someone building something. That's the best way to help us keep writing it.