Measurement system analysis (MSA)
A measurement system analysis (MSA) is conducted at the start of the Measure phase to verify that the gage (measuring device) and the people using it can detect real process variation. If the measurement system is inadequate, all subsequent data — baseline, capability, and control charts — are unreliable. An adequate measurement system is normally confirmed before the Measure tollgate.
Why MSA matters
Observed variation in data comes from two sources:
- Part-to-part (process) variation: the real differences between units
- Measurement system variation: repeatability + reproducibility errors
If measurement system variation is large relative to part-to-part variation, you cannot distinguish real process changes from measurement noise.
Gage R&R components
| Component | Definition | Detected by |
|---|---|---|
| Repeatability (EV) | Variation when the same person measures the same part multiple times with the same gage | Equipment variation: the gage itself |
| Reproducibility (AV) | Variation when different people measure the same part with the same gage | Appraiser variation: operator technique |
| Gage R&R (GRR) | Total measurement variation = repeatability + reproducibility | Combined; the sum is not simply additive in linear terms |
Acceptance criteria
| %GRR (of tolerance or total variation) | Decision |
|---|---|
| <10% | Measurement system is acceptable |
| 10–30% | Marginal; may be acceptable depending on the application and cost of improvement |
| >30% | Unacceptable; measurement system needs improvement before proceeding |
MSA study design (crossed Gage R&R)
The standard crossed Gage R&R study uses:
- 10 parts selected to represent the full range of process variation
- 2–3 appraisers (operators who normally use the gage)
- 2–3 replicates per appraiser per part
- Parts measured in random order, blind to previous measurements
Other MSA properties
| Property | Definition | How to detect |
|---|---|---|
| Bias (accuracy) | Difference between the average measured value and the true (reference) value | Measure a reference standard; compare average to known value |
| Linearity | Consistency of bias across the measurement range | Measure reference standards at multiple points; plot bias vs. reference value |
| Stability | Consistency of bias over time | Measure a reference standard on a control chart; look for drift or shifts |
| Resolution | Smallest detectable increment; gage sensitivity | Gage resolution should be ≤1/10 of the spec width or process variation |
Attribute agreement analysis
For attribute (pass/fail, go/no-go) measurement systems, use an attribute agreement analysis:
- Multiple appraisers each evaluate the same set of parts (typically 30 parts including some borderline)
- Measure: % agreement within appraiser, between appraisers, and vs. known standard
- Acceptance: typically ≥90% agreement within and between appraisers
Common exam traps
- %GRR >30% is a project stopper. You cannot trust baseline data collected with a bad measurement system; fix the measurement system before relying on the baseline.
- Repeatability is the gage; reproducibility is the operator. High repeatability error → fix or replace the gage. High reproducibility error → train operators or standardize technique.
- GRR is compared to tolerance OR total variation, depending on the standard used. Know both. The exam may specify which to use.
- NDC (number of distinct categories) ≥ 5. The NDC measures the gage’s ability to distinguish between parts. NDC = 1.41 × (part variation / GRR). If NDC < 5, the gage cannot distinguish enough categories of product.
The free 20-question CSSGB diagnostic covers all six Body of Knowledge sections, with an explanation for every answer.
Take the free CSSGB diagnosticMore: Data types · Basic statistics · Process capability · Control charts · Run charts
CSSGB Quest by Bizarre Act. Original study material; not affiliated with ASQ. CSSGB is a certification of ASQ. Terms · Privacy