You’re trying to make something better.

How do you know it’s better?

At the core of every performance metric is a belief, a fundamental belief about what makes something good.

Some things are obvious—a vacuum cleans, a flashlight lights—but most things worth measuring are anything but obvious.

Most of the time what we want to measure is impossible to measure directly, so we measure indirectly instead. Most of the time we measure performance by proxy: schools measure graduation rates, hospitals measure length of stay, businesses measure revenue, etc. But what we really want are educated students, cured patients, and satisfied customers.

How do you know your metrics measure what you want them to?

Metrics come in trees.

When you can't measure what you want directly, you start with a goal, an outcome, an end (the top of the tree) and work your way back through the branches of causes and effects (a story!) to sift out things you can measure (the leaves of the tree).

The end goal, the lagging indicator, branches out into clusters of other metrics, all hanging off each other like chains of cause and effect. The further down the tree you go, the more leading indicators you find—things that help signal success early on. But beware: the further away you get from the trunk, the less reliable those indicators become.

Every metric is a theory, a hypothesis that needs to be tested and validated by something else.

“Does this make that better?”

Does site speed improve customer satisfaction? Does customer satisfaction improve sales? At the end of the day, it often takes a judgement call—a belief in cause and effect—to decide which metrics matter most, and then data to back them up.

Proxy metrics require a separate, ultimate goal to ground them in reality.