Measurement Systems: Calibration and MSA

SEF Reference Guide G07 | Manufacturing OpEx Maturity Assessment

Supports criteria: QS-2 Measurement Systems

Glossary terms covered: Calibration, MSA

How to use this guide. This is SEF's standard method and the numbers we start from. Every operation is different, so treat the targets, timings, weightings, and templates here as a starting point and adjust them to your products, customers, and risk. Write your version down before you measure against it. Do not apply a tool just to apply a tool. Every method here carries a cost in time, attention, and discipline, and the benefit has to justify that cost in your operation. If it does not, leave it out and say why. A lower score on a criterion you have deliberately chosen not to pursue is an honest score, not a failure. One thing should not change: the maturity level definitions, because they are the scoring basis for the assessment.

What it is

Every decision about quality rests on a measurement. If the gauge is wrong, you scrap good parts, ship bad ones, and chase process problems that are really measurement problems. This guide covers the two things that make measurement data trustworthy.

Calibration answers “is the instrument accurate?” It compares the instrument against a reference standard of known value, adjusts it if needed, and records the result. The chain of comparisons must be unbroken back to national or international standards (in the US, through NIST). That property is called metrological traceability.

Measurement Systems Analysis (MSA) answers a different question: “is the whole measurement system good enough for this job?” A calibrated caliper can still give poor data if two inspectors hold it differently or the part is hard to fixture. MSA studies (Gage R&R, bias, linearity, stability, and attribute agreement) quantify how much of the variation you see comes from the measurement itself. The reference most manufacturers use is the AIAG Measurement Systems Analysis manual, 4th edition (2010).

Why it matters

Quality: A gauge that eats up a large share of the tolerance cannot tell good parts from bad ones near the limits. Escapes and false rejects both follow.
Cost: False rejects scrap good material. Chasing “process” variation that is really gauge noise wastes engineering time.
Customer trust: Automotive, aerospace, and medical customers audit calibration records and MSA. A missing certificate can hold up a shipment or a part approval.
Process control: SPC and capability numbers (see G10) are only as good as the data feeding them. Poor MSA makes Cpk meaningless.
Recall exposure: When a gauge is found out of tolerance, traceable records tell you exactly which product it measured and needs review.

Key terms

Term	Meaning
Calibration	Comparing a measuring instrument against a traceable reference standard and adjusting it so its readings are accurate. Each instrument carries a label showing its last and next calibration date.
MSA	Measurement Systems Analysis. Studies (such as Gage R&R, bias, linearity, and stability) that quantify how much of the observed variation comes from the measurement system itself, so you know whether the data can be trusted.
Metrological traceability	A documented, unbroken chain of calibrations linking a result to a reference standard, each step with a stated uncertainty.
Test uncertainty ratio (TUR)	Instrument tolerance divided by the uncertainty of the calibration. A common rule of thumb is at least 4:1.
Out-of-tolerance (OOT)	A calibration finding that the instrument was reading outside its allowed error. Triggers a review of product it measured.
Repeatability	Variation when one person measures the same part several times with the same gauge. Also called equipment variation.
Reproducibility	Variation between different people (or setups) measuring the same parts. Also called appraiser variation.
Gage R&R	A study that combines repeatability and reproducibility and expresses them as a percent of total variation or of tolerance.
ndc	Number of distinct categories. How many separate groups of parts the system can reliably distinguish.
Bias	The difference between the average measured value and the reference value.
Linearity	How bias changes across the measuring range.
Stability	How bias changes over time.
Attribute agreement analysis	The MSA study for pass/fail and visual checks, comparing inspectors to each other and to a known standard.
Process measurement instrument	An instrument built into the process, such as a pressure transducer, thermocouple, torque controller, or flow meter.

How to do it

Build the instrument register. List every device used to accept product or control the process: ID, type, location, range, tolerance, owner, calibration interval, and last and next dates. Include process instruments, not just the lab.
Set calibration rules. Use suppliers accredited to ISO/IEC 17025 for external calibration, and require certificates that state traceability and measurement uncertainty. For in-house calibration, write a procedure for each instrument type and keep your reference standards calibrated.
Label everything. Every instrument shows its ID and next due date. Anything overdue or unlabeled is taken out of use and tagged.
Set intervals from history. Start with the maker’s recommendation. Shorten the interval for instruments often found out of tolerance, and lengthen it cautiously for ones that stay stable.
Handle out-of-tolerance findings. Record the as-found reading. Assess which product was measured since the last good calibration and decide whether it needs recall or re-inspection. This is where most plants fall short.
Map measurement points. Using the control plan (see G10), confirm there is a measurement at each point where a CTQ is created, not only at final test.
Run Gage R&R on critical gauges. A typical variable study is 10 parts, 3 appraisers, 3 trials each, parts chosen to cover the real process range. Common acceptance guidance from the AIAG manual, which itself warns against using these thresholds alone (look at the charts and think about how the data will be used):
- Under 10% of total variation: acceptable.
- 10% to 30%: may be acceptable depending on the importance of the measurement and cost of improvement.
- Over 30%: not acceptable, improve the system.
- ndc of 5 or more.
Run attribute agreement on visual and go/no-go checks. Use known good, bad, and borderline parts (the boundary samples from G06). Rules of thumb in wide use: effectiveness 90% or better, miss rate 2% or less, false alarm rate 5% or less, and kappa above 0.75.
Train key people. Quality engineers and lead inspectors should be able to plan, run, and interpret MSA studies. Team leaders should know what a poor Gage R&R looks like and when to call for one.
Repeat studies on change. New gauge, new fixture, new inspectors, or a changed tolerance all call for a fresh study.

What each maturity level looks like

QS-2 Measurement Systems

Level	Assessment wording	What it looks like on the floor	How to move up
1	Calibration is ad hoc.	Gauges get checked when someone doubts a reading. No register, no labels, personal tools in use.	Build the register, pull uncontrolled gauges, and calibrate test equipment.
2	Test equipment is calibrated to a recognized standard and labeled with its last calibration date.	Lab and final test equipment carry labels and certificates traceable to national standards.	Add process instruments (torque, pressure, temperature, weigh scales) to the register and schedule.
3	As Level 2, plus process measurement instruments are calibrated on a regular schedule.	In-process instruments are labeled and on a recall system. Overdue items are rare and tagged out.	Use the control plan to check that every CTQ is measured where it is created.
4	As Level 3, plus process measurement takes place at key points in the process.	Measurements sit at the stations that create the characteristic. Data is used in real time, not just logged.	Train quality engineers and lead inspectors in MSA and run Gage R&R on critical gauges.
5	As Level 4, plus key people are trained in Measurement Systems Analysis (MSA).	Gage R&R and attribute studies exist for critical measurements, and people can explain the results. Poor systems are being improved.	Sustain: re-study on change, use results to set gauge investment, and link MSA to SPC and capability work.

Evidence to look for

Document

Instrument register covering both test and process instruments.
Calibration procedure and a sample of certificates showing traceability and uncertainty.
Out-of-tolerance records with product impact reviews.
Gage R&R and attribute agreement study reports for critical measurements.
MSA training records.

Observation

Pick ten instruments at random on the floor. Are they labeled, in date, and on the register?
Are personal or unregistered gauges in use?
Are measurements taken at the stations that create the characteristic?

Statement (questions to ask and who to ask)

Operators: “How do you know this gauge is good to use?” “What do you do if the label is out of date?”
Calibration coordinator: “What happened the last time a gauge was found out of tolerance?”
Quality engineer: “Show me the Gage R&R for this CTQ. What was the ndc? What would you do if it were 35%?”

Management report

Percent of instruments in calibration (target 100%).
Overdue calibrations by area and age.
Out-of-tolerance rate and number of product impact reviews.
Percent of CTQ gauges with an acceptable MSA study.

Common pitfalls

Calibration without MSA. A calibrated gauge is accurate against a standard. It may still be too coarse, or used too inconsistently, for the tolerance.
Ignoring out-of-tolerance results. The certificate says “found out, adjusted” and nobody asks what product it measured. That is the real risk.
Forgetting process instruments. Torque tools, thermocouples, and scales drive product quality and are often left off the list.
Bad part selection for Gage R&R. Picking 10 near-identical parts makes the gauge look bad. Picking parts wider than the process makes it look good. Sample the real range.
Skipping attribute studies. Visual inspection is often the weakest measurement system in a plant, and the least studied.

Template

Instrument Register Fields

ID	Description	Location	Range	Tolerance	Used for (test / process / CTQ)	Interval	Last cal	Next due	Calibrated by	Last result (in / OOT)	MSA status

Gage R&R Study Record

Field	Entry
Gauge ID and characteristic
Tolerance (LSL / USL)
Parts / appraisers / trials	10 / 3 / 3
Method (ANOVA or average and range)
Repeatability (% of total variation)
Reproducibility (% of total variation)
Total %GRR
%GRR to tolerance
ndc
Decision (accept / conditional / reject)
Actions and owner
Study date / next review

First 30 days

Week 1: Build or clean up the instrument register. Walk the floor and tag anything unlabeled, overdue, or personal.

Week 2: Calibrate or remove every overdue item. Write the out-of-tolerance procedure, including product impact review.

Week 3: From the control plan, list the gauges that measure CTQs. Confirm each CTQ is measured where it is created.

Week 4: Run one variable Gage R&R and one attribute agreement study on your highest-risk measurements. Share the results with the team and schedule MSA training for key people.