A capability index can look precise while describing the wrong population or an unstable process. Before comparing supplier Cpk values for a PCB feature, inspect how the data were collected and what the calculation assumes. A larger number is not automatically stronger evidence.
NIST’s process capability discussion relates capability to an in-control process and specification limits, and identifies assumptions behind common indices. The appropriate analysis depends on the data and process, not simply on a spreadsheet formula.
Identify the characteristic and population
Record the exact feature, measurement method, units, specification limits, and product revision. Clarify whether observations come from one setup, several lots, different machines, or multiple board constructions. A pooled result may conceal differences that matter to the application.
Check whether the data represent ordinary production or a selected engineering run. Sorting out nonconforming units before the analysis can make the retained distribution look better than the process that produced it. Preserve the sample-selection and exclusion rules. Ask whether repeated readings of one feature were treated as independent product observations, since that changes the meaning of the data set.
Ask for supporting evidence
| Evidence | Question answered |
|---|---|
| Time-ordered observations | Was the process stable during collection? |
| Measurement review | Is observed spread dominated by the test method? |
| Distribution and calculation method | Are the index assumptions appropriate? |
| Sample and subgroup definition | What variation does the reported estimate represent? |
Do not compare indices that use different specification limits or different definitions of variation without explaining the difference. Also distinguish a point estimate from the uncertainty associated with estimating it from a finite data set.
Turn findings into a process decision
- Verify the source data and specification revision.
- Review stability and obvious population mixtures.
- Evaluate the measurement system’s contribution.
- Confirm the statistical method with the responsible specialist.
- Document the conditions supported by the conclusion.
If a process is unstable, investigate the changing conditions before relying on one capability number to predict future output. If the process is stable but poorly centered, that is a different engineering problem from excessive spread. Retain the distinction in the corrective-action discussion.
A capability report does not replace inspection requirements or prove that every delivered board conforms. Use it as evidence about a defined process under the reviewed conditions. Revisit the analysis after meaningful process, material, or measurement changes rather than carrying an old number forward indefinitely.
The measurement-variation guide explains how repeated readings and product-to-product data help assess whether the capability report is characterizing manufacturing or the measurement system.
Move from reading to a build-ready project
Use the fabrication and assembly checklists to prepare drawings, production files and test requirements.
Open the DFM checklist