MSA Training: Measurement Systems Analysis and Gauge R&R, Online

Opens with a self-check: how would you rate your MSA knowledge, is a calibration sticker enough, and which gauge in your own plant do you quietly doubt? Then two recalls. In June 2025, Hyundai Motor America recalled 10,409 IONIQ 5 vehicles because a headlamp aiming marking, checked by eye, came out wrong after die polishing. In February 2025, Toyota recalled 40,922 vehicles after a melt mark on seatbelt webbing failed a strength test that destroys the sample. Turns these into three questions about visual checks, gauges and destructive tests. Introduces XYZ Engine Parts, a camshaft plant in Bangkok, with a brake-disc plant in Chicago and a gear plant in Frankfurt. Outlines the twelve modules in three parts, worked calculations on screen, the practice questions and the final exam, and closes with a recap.

What you will be able to do

• Explain what measurement systems analysis tells you about a number
• Recognise visual and destructive checks as measurement systems
• Identify a gauge or inspection in your own plant that needs a study

Why it matters

A gauge with a valid sticker can still mislead your quality decisions; MSA shows how far you can trust each number.

Practice: Practice: Welcome, and why measurement matters · 3 questions

Starts at XYZ Engine Parts, where a vernier caliper gives only three possible readings across camshaft journals with a forty micrometre tolerance, while a micrometer gives forty steps. Reads section 2.1.1 of the NIST/SEMATECH e-Handbook on a measurement process as a production process whose output is numbers, and its errors. Explains accuracy and precision, bias as a systematic error, and scatter made up of repeatability and reproducibility. Shows three micrometers on three machines with slightly different bias, and the master journal XYZ measures every morning on each one. Covers resolution, with a ten micrometre shank tolerance at an aerospace-fastener plant in Toronto, and why a coarse gauge cannot be studied into shape. Closes with a recap: know the errors in your numbers before you use them.

What you will be able to do

• Tell bias from repeatability and reproducibility in a set of readings
• Check that a gauge's resolution is fine enough for the tolerance
• Explain why every measured value carries part and measurement variation

Why it matters

If the gauge cannot see the differences you care about, no study can rescue it; choosing the right one comes first.

Practice: Practice: Measurement as a process · 3 questions

Opens with a XYZ micrometer that is in calibration at twenty degrees in the laboratory, yet uses about forty-five per cent of the tolerance when measuring warm journals on the line. Reads NIST handbook section 2.3.1 on calibration as comparison with reference standards to reduce bias. Explains how MSA studies the whole system: gauge, people, method, conditions and part. Covers what IATF 16949 asks: MSA for each type of measurement system in the control plan, and when one study can cover twelve micrometers of the same model. Shows an extract from the XYZ MSA plan built from the control plan, with journal diameter as a special characteristic, a study, a date and an owner, and who runs the studies. Looks at air gauges at a brake-disc plant in Chicago, and closes with a recap: calibrate the gauge; study the measurement.

What you will be able to do

• Explain the difference between calibration and measurement systems analysis
• Describe what IATF 16949 asks for MSA in the control plan
• Build an MSA plan with a study, date and owner for each system type

Why it matters

Calibration records alone do not answer an auditor's MSA questions; a plan built from the control plan does.

Practice: Practice: Calibration, MSA and the control plan · 3 questions

Starts with two grinding machines at XYZ Engine Parts: on machine four, a calibrated micrometer reads three micrometres high, so journals drift towards the lower limit unnoticed. Reads NIST handbook section 2.4.5, which defines bias as the difference between the measurement result and its true value, estimated against a reference. Shows the XYZ bias study on a reference journal of 25.000 millimetres, how to calculate the average and bias, and how the t value tests whether bias is real before judging it against the tolerance. Traces the cause on machine four to temperature and the fix of a five-minute cooling station, confirmed by a repeat study. Covers studying twelve micrometers of the same type, a bore gauge at a gear plant in Frankfurt, and closes with a recap: consistent is not the same as correct.

What you will be able to do

• Run a bias study against a reference part in normal use
• Calculate bias and test it with the t value
• Choose between removing the cause, adjusting or correcting results

Why it matters

A biased gauge does not show as scatter; it quietly moves the whole process towards a limit.

Practice: Practice: Bias · 3 questions

Opens with a XYZ micrometer that shows zero bias on one master at 25.000 millimetres but reads two micrometres out near both specification limits. Reads NIST handbook section 2.4.5.2: a linear gauge responds in equal steps, or keeps a constant bias across the range, studied with more than four reference standards. Shows how XYZ designs its linearity study with five masters from 24.980 to 25.020, measured in three rounds in a mixed order, and how to read the fitted line and test whether the slope differs from one. Traces the cause to a calibration certificate that checked only one point, then recalibration and a repeat study. Adds a medical-device moulder in Boston and a glass plant in Mumbai, and closes with a recap: check the gauge where the decisions are made.

What you will be able to do

• Design a linearity study with references across the range
• Read the fitted slope to decide whether a gauge is linear
• Decide when to calibrate, repair or correct a non-linear gauge

Why it matters

Accept or reject decisions are made near the limits, exactly where a single middle bias check cannot see.

Practice: Practice: Linearity · 3 questions

Begins with a micrometer on machine seven at XYZ Engine Parts, dropped in April and reading four micrometres high for five weeks until the customer's inspection disagreed, while machine eight's daily chart caught the same accident the next morning. Reads NIST handbook section 2.2 on statistical control of a measurement process and the check standard. Shows how XYZ sets up its stability check with three readings each morning, an average of 25.001 millimetres over a hundred days, and limits at plus or minus k times the standard deviation. Shows a signal on day one hundred and twelve, the chart of standard deviation, and drift at a paint shop in Tokyo. Covers the four steps after a signal, including re-checking parts measured since the last good point, and closes with a recap.

What you will be able to do

• Set up a check standard and a stability chart for a gauge
• Respond to an out-of-limit point with the four steps
• Recognise drift in a measurement process from its chart

Why it matters

A study is a snapshot and a gauge can change the next day; a stability chart shows when it did.

Practice: Practice: Stability: control charts for measurement · 3 questions

Opens with two studies of the same XYZ micrometer: ten consecutive journals and one skilled inspector, against journals across the process range, three inspectors and random order. Explains the idea simply, using Minitab's public guidance: repeatability as variation when one person measures the same part repeatedly, reproducibility as variation between people, compared with variation between parts. Applies the plan to the camshaft journal micrometer at XYZ, a crossed study in which every inspector measures every journal twice. Contrasts it with a test laboratory in Singapore, and covers the records to keep. Closes with a recap of the planning rules: at least ten parts from across the range, at least three operators who really use the gauge, and every part measured at least twice in random order.

What you will be able to do

• Choose at least ten parts that cover the process range
• Plan a crossed gauge R&R study with operators, trials and random order
• Explain how repeatability differs from reproducibility

Why it matters

No analysis can rescue a badly planned study; the plan decides whether the result can be trusted.

Practice: Practice: Planning a gauge R&R study · 3 questions

Starts with the XYZ study of sixty readings from ten journals, three inspectors and two trials, and the two analyses the quality team can choose. Explains the average and range method, done by hand on a data sheet, and the ANOVA method, which also tests the operator-by-part interaction. Draws on Minitab's published formulas, which cite the 2010 fourth edition of the AIAG MSA manual, and the guide to the open-source gageRR package, both requiring a balanced study. Works through both methods with the XYZ data, turning mean squares into variance components, then combining sources by adding variances and taking square roots: E-V 2.4, A-V 1.8 and gauge R&R 3.0 micrometres. Shows an interaction at an aluminium extrusion line in Dubai, and closes with a recap.

What you will be able to do

• Calculate gauge R&R with the average and range method
• Explain what the ANOVA method adds through the interaction term
• Combine variance components correctly by adding variances

Why it matters

Adding standard deviations instead of variances gives the wrong answer; working the numbers shows how the result is built.

Practice: Practice: Average and range, and ANOVA methods · 3 questions

Opens with one XYZ micrometer on two features: 45 per cent of the tolerance on a camshaft journal, but 9 per cent on a shoulder with a wider tolerance. Explains Minitab's comparisons, with the AIAG guidelines it reports: per cent study variation, per cent tolerance, per cent process and per cent contribution. Works through the XYZ numbers: repeatability 14.4, reproducibility 10.8 and gauge R&R 18.0 micrometres of study variation, and 18.0 divided by 40 giving 45 per cent. Explains N-D-C as 1.41 times the part standard deviation divided by the gauge standard deviation, truncated, with five or more as the guideline. Shows the two tests disagreeing at a bracket plant in London, explains under ten per cent as acceptable and over thirty as not, and closes with a recap.

What you will be able to do

• Calculate per cent tolerance and per cent study variation for a gauge
• Calculate the number of distinct categories and judge it
• Choose the right ratio for the job the gauge does

Why it matters

The same gauge can be fine for one feature and unfit for another; judging it against its job avoids wrong decisions.

Practice: Practice: Tolerance, process variation and ndc · 3 questions

Starts with a visual check for chatter marks on camshaft journals at XYZ Engine Parts, where a study showed one inspector accepted marked journals three times out of twenty chances. Explains the method from two public sources, the SPC for Excel guide and Minitab's guidance: parts with a reference value decided by a better method, rated several times in random order under hidden codes, and compared within appraisers, between appraisers and against the reference. Designs the XYZ study, presents its results and what the team did about them. Covers effectiveness, miss rate and false alarm rate against published guidelines, and kappa as agreement beyond chance. Shows why the reference matters at an electronics plant in Hong Kong, and closes with a recap: a good kappa can still hide a high miss rate.

What you will be able to do

• Design an attribute agreement study with known and borderline parts
• Calculate effectiveness, miss rate and false alarm rate
• Explain why a good kappa can still hide a high miss rate

Why it matters

Visual checks are measurement systems too; testing the decisions stops marked parts reaching the customer.

Practice: Practice: Attribute agreement analysis · 3 questions

Opens with the XYZ journal micrometer that failed every test in Module 9, and two ways to respond: re-run the study until it fits, or find the cause and fix it. Covers IATF 16949 clause 7.1.5.1.1 on acceptance criteria that conform to the customer's reference manuals, customer approval for other methods and keeping records of that acceptance. Shows XYZ protecting the customer with interim checks, reading the split of variation, and acting on the causes NIST lists: wear, environment, technique and resolution. Explains why averaging only reduces repeatability, how XYZ fixed both the gauge and the method, sent its results to the customer and confirmed the fix with a restudy on the same plan. Ends with a destructive study using batches of alike parts at a fastener plant in Toronto, and a recap.

What you will be able to do

• Protect the customer with interim checks when a gauge fails
• Identify the cause of measurement variation from the study split
• Plan a study for a destructive test using batches of alike parts

Why it matters

Re-running a study until it fits improves the numbers, not the gauge; fixing the cause gives an honest record.

Practice: Practice: Acting on the results · 3 questions

Begins with a question auditors often ask: pick a characteristic in the control plan and show the study for its gauge, answered in forty minutes in one plant and two minutes in another. Shows part of the XYZ MSA plan, with each measurement system, its study type and status, including the journal diameter now on a bench gauge at fifteen per cent of the tolerance with customer approval. Explains the links to SPC, to part approval through PPAP, and to measurement uncertainty in NIST handbook section 2.5. Follows the XYZ journal gauge from start to finish, then gives a road map and checklist for your own plant. Closes with a three-line recap of the course, plan, study, act and sustain, and the final exam with its certificate.

What you will be able to do

• Build an MSA plan tied to every control plan characteristic
• Explain how MSA links to SPC, part approval and measurement uncertainty
• Use a road map to start an MSA programme in your own plant

Why it matters

A programme turns individual studies into evidence an auditor or customer can follow in minutes.

Practice: Practice: Running an MSA programme · 3 questions