Measurement of 10.048, Upper Specification Limit 10.050: Is This Batch of Parts Conforming or Nonconforming? — A Five-Step Method for Selecting Measurement Uncertainty Judgment Rules

In the inspection records of a machine processing plant, a part with an outer diameter of 10.048 mm was marked as "conforming," with the specification being φ10.00 ± 0.05. Three months later, when the customer retested the same batch, a measurement of 10.052 mm was obtained at the same point, leading to a nonconforming judgment and a full batch return. The main issue is not about whose measuring tool is better—both parties used calibrated and qualified tools. The difference lies in the 0.004 mm range, which is precisely the domain of measurement uncertainty. The real problem is that the judgment rule was never clearly defined, leaving the question of what to do when "stuck in this position" unanswered.

Integrating uncertainty into judgment is not about complicating simple issues but about transforming post-measurement disputes into a reproducible rule. Here’s how to proceed in five steps.

1. The Root of the Problem: Simple Judgment is "Shared Risk"

The simplest way to make a judgment is to compare numbers: if the measurement value does not exceed the specification limit, it is judged as conforming. This is called simple judgment in ILAC-G8, where the acceptance limit equals the tolerance limit (AL = TL). The problem with this approach is the unequal risk distribution—when the measurement result is exactly on the tolerance limit, half of the measurement distribution falls outside the specification, meaning the probability of a nonconforming part being judged as conforming can be as high as 50%. Judging this batch of parts as conforming does not necessarily indicate a poor measuring tool; rather, it suggests that the rule implicitly assigns the risk to the customer.

This becomes clearer when considering the approach outlined in ISO 14253-1: a part can only be confidently judged as conforming if the measurement result, plus or minus the uncertainty, still falls within the specification. Conversely, it is judged as nonconforming. The area between these two conditions is the "uncertainty zone." Simple judgment effectively merges this uncertainty zone into the conforming zone—this is where disputes arise.

2. Step One: Calculate TUR to Determine if the Measuring Tool is Suitable for Judgment

Before selecting a judgment rule, calculate the Test Uncertainty Ratio (TUR): TUR = Tolerance Band Width ÷ Measurement Uncertainty U. This answers a prerequisite question—whether the measurement capability matches the tolerance.

A common industry threshold is TUR ≥ 4:1 (some older documents specify 10:1, but the criteria need to be aligned: whether the denominator is U or 2U). Below this threshold, simple judgment cannot support release judgments and must be supplemented with guard bands or a different measuring tool. The source of U must be clarified: the U on the calibration certificate (usually k = 2) is the inherent uncertainty of the tool, which must be combined with uncertainties from personnel, clamping, environment, and repeated measurements, using the GR&R result or re-evaluating.

Example: Specification φ10.00 ± 0.05, tolerance band T = 0.100 mm, measurement uncertainty U = 0.008 mm, then TUR = 12.5, indicating sufficient measurement capability, allowing flexibility in choosing the judgment rule. If U = 0.030 mm, TUR is only 3.3, significantly reducing the options—this is not something a judgment rule can fix.

3. Step Two: Choose One of Three Rules

Choosing a rule involves three considerations: whether the customer/contract has specified a rule (if so, follow it); the consequences of misjudging this batch (if severe, choose the three-state guard band to clearly separate the uncertainty zone); and whether both parties are willing to share the risk (if so, simple judgment can be used). Note especially: the rule must be agreed upon before measurement; selecting a rule after measurement is equivalent to fitting the data to the conclusion.

4. Step Three: Determine the Guard Band Width w

The guard band width is typically set as w = rU, with r = 1 being the most common approach, meaning the expanded uncertainty with k = 2 is used directly as the guard band. For example, with U = 0.008, the upper acceptance limit is reduced from 10.050 to 10.042: even if the true value is exactly 10.050, the probability of the measurement result falling below 10.042 is only about 2% (Φ(−2) ≈ 2.3%), reducing the false acceptance risk from a maximum of 50% to the 2% range. The trade-off is an increased false rejection probability—this is the essence of the guard band, a risk redistribution.

Some empirical values: when TUR is much greater than 4 and the customer has no objections, w can be 0, and simple judgment is sufficient; when TUR < 4, for single-sided limits, or in projects with a rich history of customer claims, w should be at least U, and 2U or a target false acceptance probability can be used to calculate r; regardless of the value chosen, the basis should be noted in the documents.

5. Step Four: Work Through an Example

Specification φ10.00 ± 0.05, USL = 10.050, LSL = 9.950. Measuring tool U = 0.008 (k = 2), TUR = 12.5. Actual measurement is 10.048.

• Simple Judgment: 10.048 < 10.050, judged as conforming—however, the difference is only 0.002, and no one can prove that the true value is within the specification.
• Three-State Guard Band, w = U = 0.008: conforming line at 10.042, nonconforming line at 10.058. 10.048 falls between these two lines, judged as "pending."
• Handling the pending status: retest with a measuring tool of smaller uncertainty. Using a device with U = 0.003, the conforming line becomes 10.047. If the retest result is 10.046, below the conforming line, it is judged as conforming and released; if the retest result is 10.049, it remains pending and is then processed through the concession acceptance procedure, requiring written customer approval or rework.

A red line: the number of retests, the measuring tool, and the method of value selection must be clearly defined in the rule. Repeated testing until a conforming result is obtained is not a judgment but data selection.

6. Step Five: Document and Contract the Rule

A rule that is not implemented is as good as not being defined. Where to document: the judgment section of the inspection work instruction (specifying the measurement rule, w value, and handling path for the pending zone), the judgment and reaction plan section of the control plan, the conformity statement on the calibration certificate (without declaring the judgment rule, the term "conforming" has no clear boundary), and the technical or quality agreements with customers and suppliers—using the same set of rules for incoming inspection to avoid the recurring scenario where the supplier deems the parts conforming and you do not.

7. Three Common Pitfalls

• Comparing numbers without writing the rule. This seems fine in daily operations, but when disputes arise, there is no common language, and the customer’s retest conclusion naturally takes precedence.
• Using an insufficient measuring tool and relying on the rule to compensate. When U = 0.050 and TUR = 2, applying a guard band directly results in a zero usable window—under such circumstances, the only solution is to change the measuring tool or the measurement method.
• Judging all pending zones as nonconforming. This leads to a pile of false rejections and continuous complaints from the production line. The correct handling for the pending zone is retesting, negotiation, or upgrading the judgment, not a blanket nonconforming judgment.

Edge data is never just a matter of bad luck; it merely exposes the rules that were not clearly defined. By aligning the tolerance band, uncertainty, and judgment rule, such disputes can have reproducible answers.

Edge data is not a matter of luck but of rules—first calculate TUR, then determine the judgment rule and guard band width, and document the exit path for the uncertainty zone in files and contracts.

Knowledge code: 6.2.1

Version: v20261016

Author: QTank QTank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners, helping companies continuously improve their quality capabilities.