🎯 Quick Answer

What: Your measurement system is the instrument and method you use to produce a number. Like any tool, it has its own error, and that error is baked into every figure it reports.

Why it matters: If the ruler cannot reliably tell a good part from a bad one, every chart built on it is confident fiction. A measurement problem often masquerades as a process problem.

How to apply: Before you analyze the data, check the measurement system. Have two people measure the same items (reproducibility) and one person measure twice (repeatability). If they do not agree, fix the ruler before you trust the number.

The payoff: You stop chasing swings that live in the gauge instead of the process, and the improvements you make are aimed at problems that are actually real.

Before you ask whether a number is good or bad, it is worth asking a harder question: can you trust the way you got it?

We treat measurements as facts. The gauge reads nineteen, so the number is nineteen. But every measurement is really two things added together: the thing you are trying to see, and the error in how you saw it. Two inspectors measure the same part and read it differently. The same person measures it twice and does not quite agree with themselves. The part never changed. The ruler did.

I have watched this quietly derail good teams in three industries that have never once compared notes. The pattern is identical. A number moves, the room reacts, a project spins up to fix the process, and months later it turns out the swing was never in the process at all. It was in the gauge.

A measurement problem wearing a process problem's clothes

Here is the trap. When a number misbehaves, the instinct is to go and fix the thing the number is about. Cycle time jumped, so we investigate the workflow. Defect rate climbed, so we audit the line. It rarely occurs to anyone that the number itself might be unreliable, because the number looks so much like a fact.

But a measurement is only as good as the system that produced it. If two people measuring the same thing get different answers, then some of the variation you are staring at is not in your operation at all. It is in your ruler. And you cannot improve your way out of a problem that lives in the measurement, no matter how clever the analysis downstream. You will just chase ghosts, spend real effort, and change a process that was never the problem.

The kind of wrong that hides inside a precise number

There is a particular version of this that fools smart people: false precision. A readout to three decimal places, taken with an instrument nobody calibrated. A dashboard reporting to the tenth of a percent, built on a definition two teams apply differently. The precision looks like rigor. It is often just confident error, dressed up.

I learned to distrust a number that is more precise than the way it was measured. Precision is how finely you report. Accuracy is whether you are right. They are not the same thing, and reporting a rough measurement to three decimals does not make it true. It just makes the mistake harder to question, because it looks so exact. If you have ever wrestled with defining a metric so two people apply it the same way, you already know this territory, and a clear data collection plan is where that discipline starts.

How to check the ruler before you trust it

The good news is that verifying a measurement system is not exotic. It is boring, quick, and almost always skipped. Do it before the analysis, not after.

Do that, and the data you take into the Measure phase is data you have earned the right to trust. It also protects the honest baseline you take next, because a before-and-after comparison is only as trustworthy as the ruler used for both.

If field notes like this are useful to you, I write more of them, and share the tools I build, in my newsletter.

Frequently asked questions

What is a measurement system in Lean Six Sigma?

It is the full set of instruments, methods, definitions, and people used to produce a measurement. In the Measure phase, verifying it (often through a measurement system analysis or Gage R&R) confirms that the numbers you are about to analyze can actually be trusted.

How do I know if my measurement system can be trusted?

Have two people independently measure the same items and see if they agree (reproducibility), and have one person measure the same item twice (repeatability). If either check shows meaningful disagreement, the variation is in your measurement, not your process.

What is the difference between accuracy and precision?

Accuracy is whether a measurement is close to the true value. Precision is how finely and consistently it is reported. A number can be very precise and still be wrong, which is why reporting to three decimals does not fix an uncalibrated instrument.

Why does a measurement problem look like a process problem?

Because the number looks like a fact. When it moves, teams investigate the operation the number describes, not the way it was measured. If the swing actually lives in the gauge, that investigation improves the wrong thing.

Should I check the measurement system before or after collecting data?

Before you rely on it. Verifying the measurement system early prevents you from building an entire analysis, and sometimes a whole project, on numbers that cannot support the weight you are about to put on them.