What Cpk targets should we use for flight-critical features?

No single Cpk target is universally correct for flight-critical features.

For flight-critical characteristics, many organizations use higher internal capability expectations than they do for standard dimensions, often starting around Cpk 1.33 for mature, stable processes and moving to 1.67 or higher for characteristics with very low tolerance for variation. Some programs set even tighter expectations for special or key characteristics. But that is a policy choice, not a law of physics, and it only means something if the underlying data is valid.

The practical answer is this: set the target based on risk, process maturity, and contractual or internal quality requirements, then verify that the metric is statistically meaningful for that feature and process.

What to use in practice

• Do not use one blanket Cpk number for every flight-critical feature. Different characteristics have different failure consequences, production volumes, process behaviors, and inspection methods.
• Use higher targets for higher-risk features, but define them in controlled procedures and product realization plans.
• Expect tighter thresholds only where the process is stable and measurable enough for capability analysis to be credible.
• If the process is not statistically stable, Cpk is not the right decision metric. In that case, focus first on control, variation sources, reaction plans, and containment.

Key constraints that matter more than the headline number

A reported Cpk can be misleading if any of the following are weak:

• Measurement system capability. If gage performance is poor relative to tolerance, the capability result is not trustworthy.
• Statistical control. Cpk assumes a stable process. Special causes, tool wear shifts, setup changes, and operator-to-operator variation can invalidate it.
• Data volume and sampling method. Small samples, filtered data, or inspection taken only after adjustment can overstate capability.
• Distribution shape. Non-normal or bounded data may need different treatment. A simple normal Cpk calculation may understate or overstate actual risk.
• Process type. Short-run machining, manual assembly, composite layup, heat treat, and special processes do not behave the same way.
• Tolerance strategy. A generous tolerance with a high Cpk may still be less protective than a tighter engineering control with lower apparent capability.

For flight-critical features, it is usually more defensible to combine capability targets with documented control methods such as reaction plans, first-piece verification, increased inspection, mistake-proofing where feasible, tool life controls, and clear escalation paths when drift appears.

What Cpk should not be used for

Cpk should not be treated as a substitute for engineering judgment, verification requirements, or release criteria. A high Cpk does not guarantee fit, function, safety, or audit acceptance. It also does not remove the need for traceability, approved process definitions, controlled revisions, and documented changes.

In regulated, long-lifecycle environments, this matters because capability calculations often sit across multiple systems: inspection software, MES, QMS, ERP, and sometimes spreadsheets. If those systems are not well integrated, you can end up with weak lineage between specification revision, measured result, nonconformance handling, and the exact process conditions that produced the part.

Brownfield reality

Most plants do not calculate and govern capability in one clean platform. They usually have a mix of legacy SPC tools, CMM outputs, ERP item masters, MES or traveler records, and QMS workflows for deviations and CAPA. That means your Cpk target is only as operationally useful as your data mapping and change control.

Full replacement of all quality and execution systems just to standardize capability management is usually not realistic in aerospace-grade environments. It often fails because of validation effort, qualification burden, downtime risk, integration complexity, and long equipment lifecycles. In practice, companies usually get further by defining a controlled policy for critical characteristics, validating the data path, and improving evidence traceability across the systems they already have.

A reasonable decision framework

• Classify the characteristic by consequence of failure and control intent.
• Check whether the process is stable enough for capability analysis.
• Confirm measurement system adequacy before setting or enforcing a target.
• Set differentiated internal thresholds, often with stricter expectations for higher-risk characteristics.
• Define what happens if the target is not met: containment, additional inspection, process adjustment, engineering review, or formal nonconformance workflow.
• Maintain traceability from requirement revision to measured evidence and disposition.

If you need one simple policy starting point, many organizations use 1.33 as a minimum for capable mature processes and 1.67 or higher for more critical characteristics, but only after confirming process stability and measurement validity. If those conditions are missing, the right answer is not to argue over the target. It is to fix the control system first.