The Gauge Was Lying: A Lean Six Sigma Lesson in Measurement System Analysis
A manufacturer of precision components had a defect rate that would not budge. Through 2020 and into early 2021 — already a hard stretch, with split shifts and remote quality reviews — an improvement team had tightened the process, retrained operators, and adjusted machine settings. Nothing held. Defects rose and fell with no clear cause. Only when a Black Belt insisted on a measurement system analysis before any further changes did the team find what was actually wrong. This anonymized account is a reminder of one of Lean Six Sigma's least glamorous but most important disciplines.
Why MSA comes before analysis
In the DMAIC cycle — Define, Measure, Analyze, Improve, Control — measurement system analysis lives in the Measure phase, and it exists to answer a blunt question: can you trust your data at all? If your measurement system is noisy, you cannot tell process variation from measurement variation. You end up reacting to numbers that say more about your gauge and your inspectors than about your process. Improving a process using bad data is not improvement; it is expensive guessing.
MSA breaks measurement error into pieces. The most familiar is a Gauge R&R study, which separates two contributors:
Repeatability — variation when the same person measures the same item repeatedly with the same gauge. This points to the equipment.
Reproducibility — variation when different people measure the same item. This points to method, training, or ambiguity in how the measurement is taken.
Together with bias, linearity, and stability, these tell you how much of the variation you see is real and how much is the act of measuring.
What the case revealed
When the team ran a Gauge R&R, the result was uncomfortable: a large share of the observed variation came from the measurement system, not the parts. Two inspectors reading the same component routinely disagreed beyond the tolerance the spec required. The 'defect rate' had been partly an artifact of inconsistent measurement, and every process change had been chasing noise.
They fixed the measurement first. Inspectors were retrained to a single documented method, the gauge was recalibrated, and an ambiguous inspection step was clarified so two people would read it the same way.
They re-ran the study. Only once the measurement system was acceptable — measurement variation small relative to tolerance — did the team trust the data enough to proceed.
Then they analyzed the real process. With trustworthy data, the genuine drivers of defects finally stood out, and the improvements that followed actually held.
The defect rate dropped within two cycles — not because the process had suddenly changed, but because the team could finally see it clearly.
The takeaway
Before you analyze a problem or judge an improvement, prove that your measurement system can be trusted. A modest MSA early in a project routinely saves months of chasing variation that lives in the gauge, not the process. It is the difference between fixing a problem and managing a rumour.
If your numbers do not add up and you suspect you are improving against unreliable data, XNM's strategic advisory can help you build the measurement and analysis discipline that makes improvement stick.