MSAMeasurement System Analysis
Measurement Systems Analysis (MSA) is the experimental and mathematical method of determining how much variation within a process is contributed by the measurement system itself.

On this page · 8 sections
- 1The Critical Prerequisite: Why Measurement Error Matters
- 2The Five Components of Measurement Variation
- 3Gage R&R Study Designs: Crossed, Nested, and ANOVA
- 4Acceptance Criteria: %SV, %Tolerance, and ndc
- 5Attribute Agreement Analysis: Measuring the Human Element
- 6MSA in Regulated Manufacturing: Calibration vs. MSA
- 7Common Pitfalls in MSA Execution
- 8Standardizing MSA with V5 Ultimate
How does MSA apply to your shop floor?
Pick your industry and scale — Ask V5 rewrites the definition in your context, gives a worked example, and shows what V5 does on day one.
01The Critical Prerequisite: Why Measurement Error Matters
In industrial quality control, it is common to jump straight into process capability studies (Cp and Cpk) or Statistical Process Control (SPC) charting. However, there is a fundamental rule in metrology: you cannot improve what you cannot accurately measure. Measurement Systems Analysis (MSA) is the essential first step that ensures the data you are using to make decisions is trustworthy. If the variation in your measurement system is too large, it will mask the true performance of your manufacturing process, leading to costly errors such as rejecting good product (producer's risk) or accepting defective product (consumer's risk).
Total observed variation in any process is the sum of two distinct components: the actual part-to-part variation of the manufacturing process and the variation introduced by the measurement system. The goal of MSA is to quantify the latter and prove that it represents only a small, acceptable fraction of the total variation. As the NIST/SEMATECH Engineering Statistics Handbook notes, characterizing the measurement process involves evaluating repeatability, reproducibility, and stability to ensure that the measurement system is 'in statistical control' before it is used to assess the process itself 2.
02The Five Components of Measurement Variation
A comprehensive MSA evaluates five key characteristics of a measurement system. These are broadly grouped into 'Location' variation (accuracy) and 'Width' variation (precision). Understanding these components is vital for troubleshooting a failing measurement system.
1. Bias (Accuracy)
Bias is the difference between the observed average of a set of measurements and the 'true' or master reference value. It is a measure of the systematic error. For example, if a 1.000 kg master weight is weighed ten times and the average reading is 1.005 kg, the bias is 0.005 kg. Bias is typically corrected through calibration, but MSA monitors whether that bias remains consistent over time.
2. Linearity
Linearity measures how the bias changes across the operating range of the instrument. An instrument may be perfectly accurate at the lower end of its range but show significant error at the higher end. A linearity study involves measuring several reference standards (typically five) that span the expected operating range and performing a regression analysis to determine if the bias is constant or variable.
3. Stability
Stability is the change in bias over time. Even a well-calibrated instrument can drift due to environmental factors, wear, or electronic degradation. Stability is monitored by measuring a 'check standard' at regular intervals and plotting the results on a control chart. If the chart shows a trend or a shift, the measurement system is no longer stable and requires investigation.
4. Repeatability (Precision)
Repeatability, often called equipment variation (EV), is the variation observed when the same operator measures the same part multiple times with the same instrument. It represents the inherent variation of the gauge itself under ideal, unchanging conditions. High repeatability error usually points to an issue with the instrument's design or maintenance.
5. Reproducibility (Precision)
Reproducibility, or appraiser variation (AV), is the variation in the average of measurements made by different operators using the same instrument on the same parts. It highlights inconsistencies in how different people apply the measurement method. High reproducibility error often indicates a need for better operator training or more standardized work instructions.
03Gage R&R Study Designs: Crossed, Nested, and ANOVA
The Gage R&R study is the primary tool for quantifying repeatability and reproducibility. Choosing the right study design is critical to getting meaningful results.
Crossed Gage R&R
The Crossed design is the most common and is used for non-destructive testing where the same part can be measured multiple times. In a typical setup, three operators (Appraisers) measure ten parts, with each operator measuring each part two or three times in a randomized order. This 'crossed' approach allows the analysis to isolate not only the equipment and appraiser variation but also the interaction between the two (e.g., does Operator A have more trouble with Part 5 than others?).
Nested Gage R&R
When the measurement process is destructive — such as testing the tensile strength of a wire or the chemical composition of a liquid — a Nested design is required. Since you cannot measure the same physical part twice, you must assume that parts within a small 'nest' or batch are identical. The study then compares the variation within these batches to the variation between batches and operators. This is inherently more difficult and requires high confidence in the homogeneity of the batches.
ANOVA vs. Average and Range (ARM) Methods
There are two main methods for analyzing Gage R&R data. The **Average and Range Method (ARM)** was the industry standard for decades because it is easy to calculate manually. However, it has significant limitations, most notably that it cannot detect the interaction between operators and parts. The **ANOVA (Analysis of Variance) method** is now the preferred standard in advanced manufacturing 1. ANOVA uses statistical variance components to provide a more accurate estimate of R&R and explicitly calculates the interaction term, providing deeper insight into the root cause of measurement error.
| Feature | ARM (Average and Range) | ANOVA |
|---|---|---|
| Ease of Calculation | High (Manual possible) | Low (Requires software) |
| Interaction Detection | None | Captures Operator-by-Part interaction |
| Accuracy | Approximation | Mathematically exact variance components |
| Confidence Intervals | Not typically provided | Calculates confidence intervals for variance |
04Acceptance Criteria: %SV, %Tolerance, and ndc
Once a Gage R&R study is complete, the results must be compared against established acceptance criteria to determine if the measurement system is 'capable.' The AIAG MSA Manual (4th ed.) provides three primary metrics for this evaluation.
- **% Study Variation (%SV):** This compares the total Gage R&R variation to the total variation observed in the study. It tells you how much of the 'noise' in your data is due to the measurement system. Under 10% is considered excellent; 10-30% is marginal but may be acceptable depending on the application; over 30% is generally unacceptable for process control.
- **% Tolerance (%T):** This is often more important for inspection processes. It compares the Gage R&R variation to the width of the specification tolerance. If the measurement error consumes too much of the tolerance, you will inevitably reject good parts or accept bad ones near the limits. The same 10/30 thresholds typically apply.
- **ndc (Number of Distinct Categories):** This is a measure of the 'resolution' of your measurement system relative to the process variation. It represents the number of non-overlapping groups the gauge can identify. An ndc of 5 or more is required for the system to be capable of process control. An ndc of 1 or 2 essentially means the gauge is only capable of telling you if the process is 'big' or 'small,' which is insufficient for SPC.
05Attribute Agreement Analysis: Measuring the Human Element
Many quality decisions are based on subjective or qualitative attributes — 'Pass' vs 'Fail,' 'Clean' vs 'Dirty,' or the classification of a defect type. Because there is no numerical scale, standard Gage R&R doesn't apply. Instead, we use Attribute Agreement Analysis.
A typical study involves multiple inspectors reviewing a set of samples that have a known 'true' state (the gold standard). The samples must be presented in a randomized order, and the inspectors should be 'blind' to the true state and to their previous results.
Key Statistical Metrics
- **Kappa Statistic:** This is the most common metric for attribute data. It measures the degree of agreement between inspectors (or between an inspector and the standard) while adjusting for the agreement that could happen by chance. A Kappa value above 0.70 is generally considered acceptable.
- **Effectiveness:** The simple percentage of correct decisions made by an inspector against the standard. High effectiveness is crucial for ensuring that defects are caught.
- **Miss Rate vs. False Alarm Rate:** A 'Miss' is failing to reject a bad part, while a 'False Alarm' is rejecting a good part. Understanding the balance between these two is essential for managing production costs and quality risks.
Attribute MSA is particularly critical in industries like pharmaceutical manufacturing for visual particulate inspection (USP <790>) and in medical devices for cosmetic defect classification. Without it, the 'inspector of the day' effectively becomes the specification, leading to inconsistent quality and compliance risk.
06MSA in Regulated Manufacturing: Calibration vs. MSA
In FDA-regulated industries and those governed by ISO 13485, the requirement for valid measurement systems is explicit. However, many organizations confuse calibration with MSA, which can lead to significant audit findings.
The Regulatory Distinction
Calibration is the process of comparing an instrument to a known standard to ensure its **accuracy** (bias). It is a snapshot in time. MSA is the process of evaluating the entire measurement **system** — including the instrument, the operator, the environment, and the procedure — to ensure its **precision** (repeatability and reproducibility).
FDA 21 CFR 820.72 states that equipment must be 'suitable for its intended purposes and capable of producing valid results.' ISO 13485:2016 §7.6 requires that monitoring and measurement be carried out in a manner consistent with the requirements. Simply having a calibration sticker on a gauge does not prove it is 'suitable for its intended purpose' if the operator's technique is so variable that the results are meaningless.
"Gage R&R does not measure the product — it measures the measurement, isolating how much of the variation you observe comes from the gauge and the operators rather than the parts."
07Common Pitfalls in MSA Execution
Even well-intentioned quality teams often make mistakes in their MSA programs. These pitfalls can lead to either an over-optimistic or an overly pessimistic view of measurement capability.
- **Inadequate Part Selection:** The parts used in a Gage R&R study must represent the full range of process variation. If you only pick parts that are very close to the target, the study variation will be low, and the %SV will look artificially high. Conversely, using 'out-of-spec' parts to artificially expand the range is also misleading.
- **Ignoring Stability:** Performing a Gage R&R once a year and ignoring the gauge for the other 364 days is dangerous. Environmental changes, wear, and tear can degrade a measurement system quickly.
- **Sample Size Issues:** A study with two operators and five parts is rarely enough to provide a statistically significant estimate of reproducibility. The AIAG 'Rule of 10' (10 parts, 3 operators, 3 trials) is a minimum for a reason.
- **Failing to Act on Marginal Results:** If a Gage R&R comes back at 25%, many teams simply file the report. A marginal result is a signal to investigate: Is it the operator? Is the fixture loose? Can we improve the lighting?
08Standardizing MSA with V5 Ultimate
V5 Ultimate transforms MSA from a disconnected series of spreadsheets into a central pillar of your Quality Management System (QMS). By digitizing the measurement process, the platform ensures that data integrity and compliance are built into every reading.
- **Direct Instrument Integration:** V5 captures data directly from digital calipers, scales, and spectrometers. This eliminates the 'Human Variation' of transcription and ensures that the MSA results are based on the same data that goes into the batch record.
- **Enforced Re-study Cadence:** MSA schedules are managed programmatically. If a gauge's MSA has expired, V5 can prevent the asset from being selected in an eBMR workflow, forcing compliance before production continues.
- **Centralized Statistical Analysis:** V5 provides built-in ANOVA engines for Gage R&R and Kappa calculators for attribute studies. This standardizes how results are calculated across the entire organization, eliminating 'Excel errors.'
- **Closed-Loop Quality:** If an MSA fails, V5 can trigger the opening of a Non-Conformance Report (NCR) and trigger a CAPA, ensuring that measurement issues are not just identified, but resolved.
Frequently asked questions
Q.What is the difference between repeatability and reproducibility?+
Repeatability is the variation within a single operator using the same equipment on the same part. Reproducibility is the variation between different operators using the same equipment on the same part. Together, they form the total precision of the measurement system.
Q.Why is ndc (Number of Distinct Categories) important?+
ndc tells you if your gauge has enough 'resolution' to see the variation in your process. If ndc is less than 5, your gauge is too 'blunt' to be used for effective process control or Statistical Process Control (SPC).
Q.When should I use a Nested Gage R&R design?+
Use a Nested design when your measurement process is destructive (e.g., testing the breaking strength of a pill or the chemical purity of a sample). Since you can't measure the same part twice, you must measure 'identical' parts from the same small batch.
Q.How does MSA relate to 21 CFR Part 11?+
If your MSA data is captured or stored electronically to support regulatory decisions, it must comply with 21 CFR Part 11 requirements for audit trails, e-signatures, and data security. V5 Ultimate handles this by ensuring all metrology data is immutable and attributed.
Q.Can a measurement system be accurate but not precise?+
Yes. An instrument can have zero bias (the average is exactly right) but have such high repeatability error that any individual reading is far from the truth. Conversely, a system can be highly precise (consistent) but systematically biased (wrong). A capable system must be both.
Primary sources
Further reading
- Statistical Process Control (SPC)See how MSA forms the foundation for all control chart activities.
- Process Capability (Cp/Cpk)Understand the relationship between measurement error and capability indices.
- Gage R&RA deeper look at the specific studies used to quantify precision.
- As-Found / As-Left CalibrationHow calibration data feeds into stability and bias studies.
- Acceptance Quality Limit (AQL)The impact of measurement error on batch release sampling plans.
Explore this topic
MSA sits inside 2 overlapping topic clusters in our glossary. Every neighbour is one click away.
Root-cause toolkit, SPC, capability and the rest of the QA practitioner's bench.
Want to see how MSA could fit into your own records and workflows? Explore the related V5 pages or talk to our team about what applies to your operation.
