Douglas Hubbard opens How to Measure Anything with a deceptively simple claim: measurement is a reduction in uncertainty based on observation. Not elimination of uncertainty — reduction. You don't need to measure something perfectly to measure it meaningfully.
This distinction matters enormously in HR, where "we can't measure that" is among the most frequently uttered phrases. Culture. Leadership quality. Potential. Psychological safety. The list of things HR professionals declare unmeasurable is long.
Hubbard's reframe: if you can observe it at all — even partially, even noisily — you can reduce your uncertainty about it. And reducing uncertainty about something that matters has value.
The real problem is usually definition
Most HR measurement problems are actually definition problems. "Employee engagement" is not unmeasurable. It's just frequently undefined. Different people in the same organization mean different things by it — emotional connection, discretionary effort, intent to stay, pride in the organization. These are related constructs, but they're not the same construct, and you can't build a valid measure of something you haven't specified.
The measurement process should start with a question that sounds almost philosophical: What would be different in the world if this thing were high versus low? If you can answer that — if you can describe observable implications — you have something measurable.
A practical test
Before you build a survey or pull a dataset, try this: write a one-paragraph description of what the world looks like when your construct is high, and another when it's low. Then ask: what could you observe that would help you distinguish between these two worlds?
Those observable implications are your measurement strategy.
If you can't write those paragraphs, you don't have a measurement problem yet. You have a definition problem. Solve that first.
The cost of not measuring
The alternative to imperfect measurement isn't perfect measurement — it's no measurement, which means decisions made on intuition and anecdote. The question isn't whether your measure is noisy; it's whether acting on that noisy measure produces better decisions than not measuring at all.
Usually, the answer is yes — as long as you're honest about the uncertainty and don't treat your measure as more precise than it is.
That honesty is the harder discipline. And it's the one most worth developing.