What Metrics to Actually Follow

Most dashboards measure everything and change nothing. They look impressive, they update in real time, and they quietly make you worse at your job — because a screen full of numbers is not the same as knowing which number to move.

So let me offer the only filter that's ever worked for me:

Track what you'd act on. Ignore what you'd only admire.

If a metric goes the wrong way and you wouldn't do anything differently, it isn't a metric — it's decoration. Pretty, reassuring, and useless. The hard discipline of measuring well isn't collecting more. It's having the nerve to ignore most of it.

Start by separating two families of numbers. There are vanity metrics — totals that only ever go up and make you feel good: total signups, total page views, total tasks "completed." And there are decision metrics — numbers that can go up or down and that tell you to do something: time-to-respond, percentage of work redone, how long a request waits before anyone touches it. Vanity metrics answer "are we big?" Decision metrics answer "are we good?" Only one of those questions improves the work.

A few rules I'd put on the wall:

Measure the few that move the work, not the many that move the dashboard. Three numbers you check every week beat thirty you glance at once a quarter. Attention is the scarce resource — not data.

Prefer rates and times over totals. "We shipped 100 things" hides everything. "Each thing took four days, down from six" tells you you're getting better, and roughly by how much. Totals flatter; rates inform.

Pick metrics close to the cause. A number is only useful if you can trace it to a decision you control. If it moves and you can't say why or what to do, it's too far from your hands to be worth watching.

Pair every speed metric with a quality one. Measure only speed and people rush; measure only quality and people stall. Watch them together — time-to-done and rework rate — or you'll optimize one straight into the ground.

And one trap to avoid: the moment a metric becomes a target people are judged on, they'll game it. Measure calls handled, and calls get rushed. Measure tickets closed, and tickets get closed without being solved. The fix isn't to stop measuring — it's to measure the outcome you actually want, not the activity that's easy to count.

Here's where I'd start this week. Write down the one decision you keep making on gut — who to hire, what to fix first, whether you're on track. Then ask: what single number, watched over time, would make that call for me? Track that. Put it where you'll see it. Ignore almost everything else.

Good measurement isn't a wall of charts. It's a short list you'd actually act on — and the quiet confidence that comes from steering by something real.