How to Measure Anything

Measurement: The Solution Exists

Hubbard opens against a common defeatism: the belief that some important things, quality, risk, customer satisfaction, brand value, simply cannot be measured. He argues this belief is almost always false and usually rests on misunderstanding what measurement means.

His central redefinition is that a measurement is a reduction in uncertainty, based on observation, expressed as a range with a probability. It does not require exactness. Once you accept that you only need to become less uncertain, not perfectly certain, the space of "measurable" things expands enormously.

He cites inspirations who measured seemingly impossible things with clever, simple methods: Eratosthenes estimated the Earth's circumference with shadows and geometry; Enrico Fermi taught students to estimate wildly uncertain quantities through decomposition. Ingenuity, not precision instruments, is what makes hard things measurable.

For a PM, the opening lesson is to reject the immeasurable label. Almost any factor that matters to a decision, however intangible it seems, can be measured well enough to reduce your uncertainty and improve the choice, if you approach measurement as reducing doubt rather than achieving precision.

The Clarification Chain

Why do intangibles feel immeasurable? Usually because they are vaguely defined. Hubbard's "clarification chain" cuts through this by forcing precision about what you actually mean and why you care.

The three links

• If it matters at all, it is detectable or observable: something in the world must be different depending on its amount.
• If it is detectable, it can be detected as an amount or a range, not just present or absent.
• If it can be detected as an amount, it can be measured.

The exercise of defining an intangible, asking what would be observably different if you had more or less of it, almost always reveals a measurable trail. "Employee morale" becomes turnover, absenteeism, or survey responses; "security risk" becomes frequency and cost of incidents. The intangibility dissolves under clarification.

For a PM, the takeaway is that when a factor seems immeasurable, the problem is usually definition, not measurement. Forcing a precise statement of what you mean and how it manifests turns a fuzzy intangible into something you can actually observe and quantify.

Measure Only What Is Worth Measuring

Just because you can measure something does not mean you should. Hubbard insists on measuring only what is worth the cost, and he provides a rigorous way to decide: the expected value of information, the value of reducing uncertainty on a given variable.

The logic is that reducing uncertainty is valuable only when that uncertainty actually affects a decision with real stakes. A variable you are unsure about but that would not change your choice is not worth measuring; a variable that is both highly uncertain and highly decision-relevant is where measurement pays off.

Most organizations get this exactly backwards, a pattern Hubbard calls "measurement inversion": they measure the things that are easy and comfortable, which usually turn out to be low-value, while ignoring the few high-uncertainty, high-impact variables that most influence the decision.

For a PM, the lesson is to compute, at least roughly, the value of information before investing in measurement. Directing your analytics and research effort toward the high-uncertainty, high-stakes variables, rather than the easy metrics, is how you avoid measurement inversion and get real value from measuring.

Calibrated Estimates

Before gathering new data, Hubbard shows how to extract better estimates from what people already know. The problem is that experts are systematically overconfident, their stated confidence ranges are far too narrow, so their judgments mislead.

The solution is calibration training: teaching people to give honest probability ranges, such as 90 percent confidence intervals, that are correct about as often as claimed. Through feedback and simple techniques, notoriously overconfident estimators can become well-calibrated, so their ranges genuinely capture the truth about nine times in ten.

Calibrated estimates are valuable because they let you quantify uncertainty honestly using existing expertise, no new data required. A calibrated expert's range is a real measurement, an honest reduction of uncertainty, and it forms the starting point that further data can then refine.

For a PM, the takeaway is to replace false-precision point estimates with calibrated ranges for uncertain inputs like adoption, effort, or impact, and to train yourself toward honest confidence intervals. A well-calibrated range communicates real uncertainty and is a more truthful basis for decisions than a single made-up number.

You Need Less Data Than You Think

A liberating theme is that useful measurement usually requires far less data than people assume. The instinct that you need a huge, rigorous study before you can learn anything is wrong; even a few well-chosen observations can dramatically reduce uncertainty.

Hubbard illustrates with the "Rule of Five": take just five random samples of anything, and there is a 93 percent chance that the true median lies between the smallest and largest values you observed. From near-total ignorance, five data points give you a meaningful, high-confidence range.

Combined with decomposition (breaking a hard quantity into estimable pieces, Fermi-style) and simple statistics or Monte Carlo simulation to combine uncertain inputs, this means the first few observations are often the most valuable. Uncertainty drops fastest at the start, so a small measurement frequently answers the question well enough.

For a PM, the lesson is to start measuring with small samples and quick experiments rather than waiting for a perfect dataset. Because the first handful of observations cuts uncertainty the most, a modest, fast measurement is usually enough to make a materially better decision.

A Universal Approach to Measurement

Hubbard assembles these ideas into a repeatable method, which he calls Applied Information Economics. It is a general procedure for measuring anything that matters to a decision, turning the philosophy into a workflow.

The steps run: define the decision and the variables involved; determine what you currently know, using calibrated estimates to quantify uncertainty; compute the value of information to decide what is worth measuring; then measure the high-value variables with methods appropriate to their value, and feed the results back into the decision model.

The approach unifies clarification, calibration, value of information, and efficient measurement into one loop aimed squarely at better decisions. It treats measurement not as an academic exercise but as an economic activity: spend on reducing uncertainty exactly where doing so most improves the decision.

For a PM, the takeaway is a workflow you can adopt: frame the decision, quantify your current uncertainty with calibrated ranges, identify the few variables worth measuring, and measure just enough to decide better. Measurement becomes a targeted tool for reducing the uncertainty that most affects your choices.

The Entire Book in One Framework

The whole book overturns the notion of the immeasurable by redefining measurement as the reduction of uncertainty. With the clarification chain, any factor that matters can be defined and detected; with calibrated estimates, existing knowledge becomes an honest starting range; and with the value of information, you measure only what actually affects a decision.

Because you need far less data than you think, a few well-chosen observations, decomposition, and simple simulation often suffice. Assembled into a repeatable method, this lets you measure anything worth measuring, avoiding the measurement inversion that wastes effort on easy, low-value metrics.

How to Measure Anything is not "measure everything precisely." It is a reframing: measurement means reducing uncertainty, so the real questions are what is worth measuring and what is the least you need to observe to decide better, and by those questions, almost nothing is truly immeasurable.

10 Most Important Takeaways

• Measurement is a reduction in uncertainty, not the achievement of precision.
• Most "immeasurable" things are just poorly defined; clarify what you mean.
• The clarification chain: if it matters it is detectable, and if detectable it can be measured.
• Measure only what is worth measuring, using the value of information.
• Beware measurement inversion: effort on easy, low-value metrics, neglect of high-value ones.
• Train calibrated estimators to give honest confidence ranges instead of overconfident guesses.
• You need far less data than you think.
• The Rule of Five: five samples put the true median between the min and max 93% of the time.
• Decompose hard quantities into estimable pieces, Fermi-style, and simulate to combine them.
• Follow a repeatable loop: define the decision, quantify uncertainty, measure what matters.

The deepest idea is that the belief in immeasurability is an excuse, not a fact. When we say something cannot be measured, we usually mean we have not defined it, or we assume measurement requires impossible precision. Drop both assumptions, treat measurement as reducing uncertainty about a decision, and you gain the power to quantify the very things, risk, value, quality, that most affect your choices and that everyone else leaves to guesswork.