A medical components supplier was rejecting good parts and passing marginal ones on a critical bore. A Gauge R&R study showed the measurement system used 38% of the tolerance; a fixtured air gauge cut that to 9% and false rejects from 4.2% to 0.6%.
Key numbers
A medical components supplier had a frustrating problem. Its critical bore diameter seemed impossible to control. Rejects ran above 4%, process changes made no lasting difference, and some parts that had been accepted were later found out of tolerance at the customer. The process engineers were convinced the moulding process was unstable. The process data did not quite support that.
The situation
The bore was critical to the device's function, with a tight tolerance and a customer who expected capability evidence. Measurement was done on the line with a hand-held bore gauge by three inspectors across three shifts.
What the data showed
Before any further process work, the team ran a Gauge R&R study: ten parts selected across the normal range of the process, three appraisers and three repeat measurements of each part, in random order.
The measurement system consumed 38% of the tolerance, well above the 30% level generally regarded as unacceptable.
Only 62% of the measured variation came from real differences between parts. Repeatability, the gauge itself, contributed 23%, and reproducibility, differences between appraisers, contributed 15%. Watching the inspectors explained why: each positioned the part slightly differently in the hand-held gauge, so the same part could pass for one inspector and fail for another.
The process had never been as bad as the data suggested, and improvement projects had been chasing measurement noise. It is exactly the trap a structured Six Sigma approach, of the kind we described when launching our Lean, Six Sigma and Kaizen services, is designed to catch: prove the measurement before you trust the data.
What changed
- A fixtured air gauge replaced the hand-held bore gauge, locating the part consistently and removing technique from the measurement.
- The measurement method standardised with a visual work instruction.
- Appraisers trained on the new method, and the study repeated to confirm the improvement.
Results
%GRR fell from 38% to 9%. False rejects dropped from 4.2% to 0.6%. Measurement time fell from 95 to 40 seconds per part, because the fixtured gauge was faster as well as more precise. Part-to-part variation now accounted for 91% of what was measured.
Most importantly, later Six Sigma work on the process itself became possible, because the data could finally be trusted.
Lessons for your operation
- If process improvements are not working, check the measurement system before the process.
- A Gauge R&R study takes an hour or two and can save months of misdirected effort.
- High reproducibility error points to method and technique; fixtures and clear instructions fix it.
- False rejects cost money too: good parts scrapped, and inspection time wasted.
Every critical measurement that drives decisions deserves a measurement system analysis.
