Is 0.1 always the acceptance limit?

No; it is a commonly cited goal, not a universal rule.

Does a smaller tolerance improve the ratio?

No. With fixed measurement spread, narrowing tolerance increases P/T.

Can averaging repeated measurements reduce it?

Possibly for independent random components, but averaging does not automatically remove bias or shared sources of variation.

AI Glossary / Statistics and probability / Intermediate

P/T ratio

Definition

The P/T ratio, or precision-to-tolerance ratio, compares a stated width of measurement-system variation with the permitted specification width. Under a common six-sigma convention, P/T = 6 sigma_m/(USL - LSL), where sigma_m is the relevant measurement-error standard deviation and USL and LSL are the upper and lower specification limits.

What it means

The ratio asks how much of the allowed tolerance range is occupied by measurement noise. A smaller value means that the specified measurement spread is smaller relative to the tolerance, under the chosen convention and study conditions.
It is a dimensionless summary of measurement precision relative to requirements. It is not the probability of accepting a defective item, a complete uncertainty statement, or a measure of the production process's own capability. A precise instrument can still be biased.

How it works

Define the characteristic, units, specification limits, and measurement procedure. Confirm that the tolerance is a legitimate engineering or customer requirement rather than a range selected to make the instrument look favorable.
Estimate measurement-system variation from an appropriate study. Repeated measurements can characterize repeatability, while operator, day, setup, or instrument differences may contribute reproducibility components. The components included must match intended use.
When independent variance components are combined, add their variances and take the square root. Adding their standard deviations directly would describe a different calculation. Dependencies, interactions, and study design can require a more detailed model.
Multiply the resulting standard deviation by the agreed spread factor and divide by tolerance width. NIST's glossary describes three standard deviations divided by the corresponding half-tolerance. For a two-sided tolerance interval this equals six standard deviations divided by its full width.
State the multiplier explicitly. Reports using different spread factors are not directly comparable without adjustment. Under a normal error model, a six-sigma width corresponds to the interval from minus three to plus three standard deviations, but the ratio itself does not establish normality.
Assess bias, stability, resolution, and other uncertainty sources separately. The NIST glossary mentions 0.1 as a common development goal, but a threshold is not a universal acceptance rule. The consequences of measurement decisions and the applicable requirements determine what performance is adequate.

Three practical examples

Dimensional inspection: engineers compare a gauge's measured variation with the allowed width range of a manufactured component.
Automated vision: a team evaluates whether repeated AI-assisted dimension measurements are sufficiently precise relative to the product tolerance.
Supplier qualification: reviewers compare measurement-study results only after aligning tolerance definitions, variation components, and spread multipliers.

Think of it this way

Think of drawing a fuzzy line inside a permitted band. P/T compares the line's stated fuzziness with the band's width. A narrow fuzzy line can still be drawn in the wrong place, which is why precision does not eliminate the need to examine bias.

Worked example

Suppose a fictional part has lower and upper specification limits of 9.90 mm and 10.10 mm. The tolerance width is 0.20 mm. A measurement study estimates independent repeatability and reproducibility standard deviations of 0.003 mm and 0.004 mm.
The combined measurement standard deviation is sqrt(0.003 squared + 0.004 squared) = 0.005 mm. Its six-sigma width is 6 times 0.005 = 0.030 mm. Therefore P/T = 0.030/0.20 = 0.15, or 15%.
Using NIST's equivalent half-width convention gives three sigma = 0.015 mm and half-tolerance = 0.10 mm. Their ratio is again 0.15.
If the tolerance width is tightened to 0.10 mm while measurement performance remains unchanged, P/T becomes 0.030/0.10 = 0.30, or 30%. The instrument did not become noisier; the requirement became more demanding.
A constant measurement bias of 0.020 mm would not change the calculated standard deviation or this precision-only ratio. It could nevertheless materially affect decisions near a specification limit. The example therefore cannot establish overall fitness for acceptance testing from P/T alone.

Illustrative case study

Illustrative fictional case: Aspen Components introduces an AI-assisted camera measurement system. Its team initially reports only short-term repeatability from one operator and one lighting setup.
A wider study includes normal setup changes and different operators, revealing additional measurement variation. The team updates the stated P/T basis and separately checks bias against reference measurements. It evaluates performance under the actual product tolerance and decision procedure before deployment. This is a fictional illustration, not evidence of a particular camera's performance.

Common misconceptions and mistakes

Do not substitute the production process standard deviation for measurement-system error without explaining the change in meaning. Do not mix half-tolerance and full spread, or full tolerance and half spread.
Avoid presenting 15% as a 15% defect rate. Do not treat a small ratio as proof of accuracy, traceability, or negligible classification risk. A single study can also miss longer-term drift or variation absent from its design.

Business use

P/T helps businesses judge whether measurement precision is proportionate to a specified tolerance and prioritize improvements to inspection systems. Useful reports include the study design, included variation components, uncertainty in their estimates, spread convention, and separate bias assessment. Decisions should reflect the intended use and the cost of wrong acceptance or rejection.

Check your understanding

Why is the combined standard deviation 0.005 mm?

Answer: The assumed independent variances, not the standard deviations, are added.

Why does the ratio double when tolerance width halves?

Answer: Its numerator stays fixed while its denominator is reduced by half.

Related terms

Precision
Bias
Tolerance
Specification limits
Measurement system analysis
gauge repeatability and reproducibility
measurement uncertainty
process capability

Learn next

Next study the design of gauge studies and variance-component estimation. Then examine how measurement uncertainty, bias, and decision rules affect classification near a specification boundary.

Uncertainty Bias

Sources

NIST/SEMATECH, Engineering Statistics Handbook glossary, P/T ratio: https://www.itl.nist.gov/div898/handbook/glossary.htm — three-sigma/half-tolerance definition and contextual development goal. NIST/SEMATECH, Gauge R & R studies: https://www.itl.nist.gov/div898/handbook/mpc/section4/mpc4.htm — measurement performance components. NIST/SEMATECH, Statistical control of a measurement process: https://www.itl.nist.gov/div898/handbook/mpc/section2/mpc2.htm — bias and variability over time. Numerical values and the case are original.

Previous: P-rep Next: P-value