More than once I’ve found myself in front of a dashboard where everything was green, inside a company that was doing badly. Nobody had tampered with anything. The numbers were correct, well collected, refreshed every night.
They were just measuring the wrong thing, and doing it with great precision.
A metric never lies. That’s what makes it dangerous. It tells you exactly what it measured, and if you chose badly what to measure, it tells you a truth with no connection to the question you were asking. The trouble isn’t in the data. It’s that after a few months of looking at it, you start believing it more than you believe what you can see.
Then the second thing happens, which is worse.
When a number becomes the way work gets judged, people get extremely good at making it go up. Almost never out of dishonesty. Simply because it’s the only clear signal they receive about what matters, and nobody works against the only clear signal they receive.
I watched a support team measured on first response time improve that time by forty per cent in a quarter. They’d managed it by sending everyone an immediate holding reply and dealing with the actual problem afterwards. The number was beautiful. Clients were angrier than before, because now they got two messages instead of one and the second still arrived late.
Nobody cheated. They did exactly what had been asked.
The point that interests me isn’t that you should pick better metrics, which is advice both true and useless. It’s that every metric, the moment it becomes a target, stops describing reality and starts shaping it. From then on you aren’t observing. You’re instructing.
And if you’re instructing, the question is no longer is this number right. It’s a different one. If someone made this number go up in the stupidest way possible, would the business be better or worse off?
It’s an unpleasant question, because you ask it with people in the room and it sounds like an accusation. But it’s the only way I know to find out in advance what kind of behaviour you’ve hung on that figure. The answers come back fast, because usually somebody had already thought of it.
Then there’s a whole category of numbers that measure nothing but console, and they’d deserve their own piece. Contacts collected without knowing how many were the right ones. Impressions. Hours worked. These are metrics that rise with effort, which is exactly why they’re liked: they give back the feeling of having done something, which in hard periods is a valuable commodity.
The cost of keeping them isn’t that they’re useless. It’s that they occupy the space where the useful ones would go, and every dashboard has a limited mental capacity before people stop looking at it.
The most useful thing I’ve seen people in this situation do isn’t adding indicators. It’s removing them. Staying with three numbers you’re willing to answer for in front of someone, and accepting that everything else stays dark for a while.
The dark is less frightening than a bright green light on something you weren’t watching.