The Measured and the Important
Every article on this site rests on the same observation, so it is worth stating once on its own.
What gets measured is what is cheap to measure. What matters is frequently expensive to measure. These are not the same set, and nothing automatically reconciles them.
Why the sets diverge
Instrumentation decides availability. A badge reader produces attendance data whether or not anybody wanted it. A conversation about whether somebody can see a future here produces nothing unless somebody holds it and writes it down.
Availability decides use. A number in a system gets into a report; a number that requires effort does not.
And use decides importance, backwards. Once attendance is in the monthly report it acquires weight, and the weight comes from its presence rather than from evidence that it predicts anything.
No individual decision in that chain is wrong. The outcome is a management practice built around whatever the tooling happened to emit.
The three failures this produces
Optimising the proxy. Activity scores rise while work does not, because people respond to what is measured.
Ignoring the unmeasured. Manager quality moves turnover by a factor of four and appears in no automated report.
And false confidence. A dashboard full of precise figures feels like understanding, and precision is not accuracy — which is the image this whole site is built on.
What actually fixes it
Measuring a few important things slowly, rather than many easy things continuously.
Turnover grouped by manager. Exit reasons recorded the same way each time. Whether problems surface before they become resignations.
These take a person's time and produce a handful of numbers a year, which is why they lose to a dashboard that produces thousands.
And they are the ones that move with the outcome, which is checkable on your own data rather than taken from this site.
The honest limit
Some important things are not measurable at all, and pretending otherwise produces a worse metric rather than a better decision.
These are known by paying attention, and the correct response to an unmeasurable but important thing is to look at it rather than to invent a number for it.
Why this is not an argument against measurement
Measurement is how you find out you were wrong.
The objection is to measurement chosen by convenience and then treated as the definition of performance. The same distinction applies to tools, to careers, and to every other section here.
Measure deliberately, few things, and check that they move with what you care about. That is the entire method.
The short version
- What gets measured is what is cheap to measure; what matters is frequently expensive, and nothing reconciles the two automatically
- Instrumentation decides availability, availability decides use, and use confers importance backwards
- Three failures: optimising the proxy, ignoring the unmeasured, and false confidence from precise numbers
- The fix is a few important measurements taken slowly, which lose to dashboards because they produce a handful of numbers a year
- Some important things are not measurable, and the right response is to look at them rather than invent a metric
- This is not an argument against measurement but against measurement chosen by convenience and then treated as the definition of performance
For additional context on this topic, see Minecraft.