MSA Method 2 ndc Calculation and Audit Troubleshooting

Scope and evidence limits

MSA Method 2 is described here as a Gage R&R study that estimates repeatability, reproducibility, and, when included by the selected analysis, interaction. The available evidence identifies the number of distinct categories (ndc) as an additional measurement-system metric and reports an acceptance target of ndc >= 5 in AIAG MSA fourth-edition usage.

The evidence contains an older recollection that ndc should be at least 4, but the more detailed fourth-edition account states a minimum of 5. Treat 5 as the documented decision threshold for the cited fourth-edition method. Verify customer-specific, corporate, and current manual requirements before releasing the study.

What ndc measures

ndc estimates how many non-overlapping measurement categories the system can distinguish across the observed part variation. The category width is governed by Gage R&R variation, while the available span is governed by part variation. A larger part-to-measurement variation ratio therefore produces more distinguishable categories.

Do not equate ndc directly with display resolution. Display discrimination can influence measurement variation and therefore affect ndc indirectly, but ndc evaluates categories across part variation rather than merely counting display increments. The evidence also rejects interpreting the factor in the calculation as a 97% confidence coefficient; it associates that factor with the variance relationship used in the analysis.

Calculate ndc in an MSA Method 2 worksheet

Use the part-variation and Gage R&R outputs from the same study and calculation convention:

ndc_raw = 1.41 * (PV / GRR)
ndc = floor(ndc_raw), when ndc_raw > 1

PV is part variation and GRR is total Gage R&R variation. One source writes the denominator as RR; the detailed evidence describes the intended quantity as GRR. Confirm that an existing worksheet uses compatible variation units and does not mix variances, standard deviations, percentages, or tolerance-based values.

The fourth-edition description requires rounding down to a whole number when the calculated value exceeds 1. Ordinary rounding can incorrectly promote a marginal result into the acceptable range.

Relate ndc, GRR, PV, and total variation

The evidence defines Gage R&R as the quadratic combination of repeatability, reproducibility, and interaction where interaction is included:

GRR = sqrt(EV^2 + AV^2 + IA^2)

EV denotes repeatability, AV reproducibility, and IA interaction. It also defines observed total variation as the quadratic combination of actual part variation and measurement-system variation:

TV = sqrt(PV^2 + GRR^2)

These relationships explain why ndc is sensitive to the parts selected for the study. If sampled parts cover too little actual process variation, PV falls and ndc can fail even when the instrument itself has not changed. Conversely, a broad spread of parts can increase ndc without proving that every product result conforms to specification.

Separate discrimination from tolerance conformity

This separation resolves the apparently contradictory cases in the evidence. A digital projector can return measurements inside tolerance while producing ndc < 5; conformity of the sampled parts does not establish adequate discrimination. An analog projector can produce sufficient ndc while some measured values lie outside tolerance; broad part variation can improve category separation, but it does not make nonconforming parts conforming.

Correct an Excel template that omits ndc

1. Confirm that the worksheet produces PV and total GRR from the same Method 2 study.
2. Check whether interaction is included. If it is, verify that IA contributes through the quadratic GRR equation rather than by direct addition.
3. Add ndc_raw = 1.41 * (PV / GRR).
4. Implement downward whole-number truncation for results greater than 1.
5. Display the raw ratio, reported integer, applicable threshold, and pass/fail decision separately so an auditor can reproduce the result.
6. Cross-check the worksheet against statistical software that reports ndc, using identical observations, study structure, and variation conventions.

Do not copy a formula into a worksheet until the meaning and units of its existing PV, RR, and GRR fields are known. Matching labels do not guarantee matching statistical definitions.

Diagnose ndc below 5

1. Recalculate the raw PV/GRR ratio and confirm that the reported value was truncated rather than conventionally rounded.
2. Verify that PV and GRR use compatible units and derive from the same data set.
3. Review part selection. Determine whether the sample represents the intended process variation instead of a narrow cluster of similar parts or five tolerance-position labels chosen in advance.
4. Review repeatability, reproducibility, and interaction contributions to identify which component increases GRR.
5. Keep the ndc decision separate from tolerance results. In-tolerance readings do not cure inadequate category discrimination, and out-of-tolerance parts do not invalidate an otherwise reproducible discrimination calculation.
6. Record the intended use of the measurement system and the applicable customer or corporate acceptance rule before declaring it capable or incapable.

The evidence mentions an interpretation that five categories correspond to positions below the lower limit, below tolerance center, at center, above center, and above the upper limit. That interpretation conflicts with the detailed definition based on non-overlapping GRR-width categories spanning observed PV. Do not preselect five tolerance zones as a substitute for calculating ndc from study variation.

Prepare an audit-ready record

Retain the study observations, part-selection rationale, operator and trial structure, analysis method, component variation outputs, raw ndc calculation, truncation rule, and final decision. State explicitly whether the assessment concerns process discrimination, product conformity, or both.

If an auditor requests ANOVA, document whether the selected analysis estimates interaction and ensure the worksheet applies that model consistently. The evidence reports ANOVA being requested in an audit, but it does not establish ANOVA as a universal prerequisite. Resolve the controlling requirement from the applicable customer, corporate, certification, and current reference documents rather than treating one audit request as a general rule.

FAQ