The best wrong number

Numbers look exact. Fourteen percent. Two point three million. Sixty-three days. Hard edges, no wobble. That’s the trick of them.

Most are approximations in a costume.

Pick a KPI, any KPI. You name the number, the team chases it, the number moves. Up if you wanted up. Down if you wanted down. Works nearly every time. That’s not evidence the number was right. It’s evidence that people do what you ask.

Sales are up thirty percent this quarter. Wonderful. Some of that is real growth, and some of it is a rep promising whatever it took to hit the number — a promise nobody in engineering ever agreed to. You book the revenue in ninety days and you pay for the promise for two years. The number was honest about the quarter and silent about everything after it.

Same thing on our side of the house now. AI adoption rate. Tokens burned. Share of code written by an agent. All real, all measurable, all a slice. A team where agents write sixty percent of the code might be flying. Or drowning in code nobody left can explain. The ratio doesn’t know the difference.

That’s fine.

Because the alternative to a rough number isn’t a precise one. It’s a shrug.

GDP was stitched together in the 1930s, and the man who built it told Congress, in writing, that you can’t read a nation’s welfare off of it. He was right. It counts the hurricane cleanup and skips the parent raising a kid. Every country on earth still uses it to decide whether it’s in a recession.

The S&P 500 is five hundred companies standing in for an entire economy, weighted so the ten largest carry better than a third of it. It isn’t the economy. It isn’t even the stock market. It’s still the first number anyone checks in the morning.

Retail has sales per square foot. It says nothing about whether the store is a place anyone wants to stand in, or whether the person who bought something ever comes back. You can lift it by cramming in racks and tearing out the seating. Every operator and every landlord still knows theirs by heart.

Wrong in the details, all three. Nobody’s giving them up.

Here’s what a rough number does that a good instinct can’t.

It sets a baseline. Before the number, “how are we doing” is a conversation about feelings, and the loudest person in the room wins it. After the number there’s a mark on the wall. You don’t have to trust the mark. You just have to stand next to it.

It shows a trend. One reading is noise. Twelve in a row are a direction. The value can be wrong every single time and the slope still tells you the truth — as long as it’s wrong the same way each time.

And it points. A number says this, not that. Out of the hundred things a team could care about, here are the two we’ve picked. That isn’t measurement. That’s a decision, and everyone can see where it’s aimed.

A bathroom scale can’t tell muscle from water, doesn’t know you slept badly, and reads a pound off depending on where you set it down. You step on it every morning anyway. Not because it’s accurate. Because it’s there, and it’s the same kind of wrong every time.

Steady wrong beats vague right.