6.1 Measurement System Analysis (MSA / Gauge R&R)

Before measuringtrusting parts,data, weyou must measuretrust the tool that generated it. If your ruler is elastic, every measurement system.is a lie. Measurement System Analysis (MSA) determinesquantifies ifthe error introduced by the gauge and the operator. It answers the critical question: "Is the variation inI datasee comes from the part (actual)real, or theis gaugeit (error).just noise in my test equipment?" If your measurement error exceeds 30%, you are effectively flipping a coin to determine Pass/Fail.

Gauge R&R (Repeatability & Reproducibility)

• We decompose measurement error into two distinct vectors. You must isolate them to fix them.

  Repeatability (Equipment Variation):

  The inherent precision of the hardware.

  • Test: CanOne operator measures the same operator using the same gauge measure the same part 10 times and get the same result?times.
  • If variation is high → Then the fixture is loose, the sensor is noisy, or the part is not seated consistently.

  Reproducibility (Appraiser Variation):

  The consistency between human operators.

  • Test: Can threeThree different operators measure the same part.same
  • If partvariation andis gethigh → Then the sameSOP result?is ambiguous, or the reading method is subjective (e.g., "visual estimation").

  Pro-Tip: Always perform Gauge R&R using parts that span the full tolerance range (Low, Nominal, High). If you only measure "perfect" parts, you will calculate a false sense of security.

  AIAGThe Logic of Acceptance Criteria(AIAG Standard)

  Do not debate these numbers. They are the industry standard (AIAG MSA Manual).

  Error Band Logic:

  • If Gauge R&R (GRR) < 10% Error:→ Then System is Acceptable.Capable. No action required.
  • If The measurement systemGRR is robust.
  • 10% – 30% Error:→ Then System is Marginal.Marginal.
    • Action: MayAcceptable beonly acceptablefor basednon-critical ondimensions applicationOR importanceif andthe cost of gauge.a better gauge is prohibitive. Requires Quality EngineeringManager approval.sign-off.
  • If GRR > 30% → Then System is Unacceptable.
    • Action: Stop Use. Any data collected is statistically invalid. Repair the fixture or retrain operators immediately.

  Number of Distinct Categories (ndc)

  Precision is useless without resolution. The ndc metric tells you how many "buckets" your ruler has within the process range.

  Resolution Rule:

  • If ndc < 5 → Then the gauge is too dull to detect process changes. It is like measuring a hair's width with a yardstick.
  • > 30% Error:Action: Unacceptable.Switch Theto systema cannotgauge distinguishwith betweenhigher goodresolution and(e.g., bad parts. Datamove from thiscalipers toolto is invalid. Corrective action (fixture repair, training) is mandatory.micrometers).

  Final Checklist

  MetricControl Point	AcceptanceCritical LimitRequirement	ActionRisk for FailureAvoided
  %GRR GRR< 10%	<Green 10%Light. (Green)Measurement system is robust.	ReleaseFalse toRejects Production/ False Accepts
  %GRR GRR> 30%	Red Light.10-30% (Yellow)Stop using the gauge immediately.	ConditionalProducing Releasescrap (Monitor)
  %without GRR	>knowing 30% (Red)	Quarantine Tool / Redesignit
  ndc	>Must be ≥ 5.	IncreaseInability gaugeto resolutiontrack process drift
  Sample Size	Min 10 parts, 3 operators, 3 trials (Total 90 data points).	Insufficient statistical power
  Calibration	MSA is NOT calibration. Calibration sets the zero; MSA validates the variance.	Confusing accuracy with precision