A measurement system that consumes a third of your tolerance turns a capable process into a failing one on paper. This is a working guide to deciding whether a gauge belongs in your control plan.
Start from the tolerance
Resolution should be no worse than a tenth of the tolerance being measured. Below that, the gauge cannot see the variation you are trying to control, and the chart goes flat in a way that looks reassuring and tells you nothing.
Run the study on the real process
- Real production parts that cover the real range, instead of masters or a hand-picked spread.
- The operators who will actually take the measurement.
- The environment it will live in, including the temperature swing across a shift.
Read the study properly
Gauge R&R under ten per cent of tolerance is acceptable. Between ten and thirty per cent it depends on cost and criticality, and above thirty per cent the system is not fit for the decision you are asking it to make. Look at reproducibility and repeatability separately, because an operator spread and a worn anvil need different fixes.
Plan for drift
Decide the calibration interval, what happens when a check fails, and who is allowed to adjust the gauge. Without a reaction plan, every decision the gauge feeds rests on an input nobody controls.
This is a sample post. It shows how a Guide is laid out. Delete it or replace it once there is real material to publish.