A supplier presents an impressive capability index. The receiving company accepts it without checking the process, sample or measurement method. Later, delivered parts fail the requirement.
Capability analysis compares process behavior with specifications. Its usefulness depends on the conditions behind the number and the decision the business is trying to make.
Establish the foundation
Confirm the characteristic, specification, measurement method and operating scope. Evaluate whether the data represents a stable process and the conditions under which future output will be produced.
NIST describes capability as a comparison between an in-control process and its specification limits. Stability is a foundation for interpreting the index as evidence about predictable process behavior.
An index calculated from changing conditions may summarize the sample without providing a reliable basis for the intended conclusion.
Distinguish spread from centering
For a two-sided characteristic, Cp compares specification width with six estimated standard deviations. Cpk also considers how close the process mean is to either specification limit.
| Hypothetical characteristic | Value |
|---|---|
| Lower specification limit | 9.5 |
| Upper specification limit | 10.5 |
| Process mean | 10.2 |
| Estimated standard deviation | 0.1 |
| Cp = (10.5 − 9.5) / (6 × 0.1) | 1.67 |
| Cpk = (10.5 − 10.2) / (3 × 0.1) | 1.00 |
Cpk uses the upper limit because it is nearer the mean. The example shows how a favorable spread can coexist with poor centering. NIST provides these formulas and explains their distinction.
Check what the sample represents
Record the machines, tools, operators, material lots and period included. A controlled short trial may differ from sustained production.
Do not combine data merely to increase sample size. A mixture of different processes can obscure the behavior of each.
A hypothetical supplier demonstrates capability using one carefully adjusted machine. The commercial plan then distributes production across several machines. The original evidence needs to be assessed against the actual delivery arrangement.
Verify the measurement system
If measurement variation is substantial relative to the characteristic, the observed distribution can misrepresent production behavior.
Check calibration, method consistency, resolution and relevant reproducibility. Include the people and conditions used in routine inspection.
A more detailed spreadsheet cannot compensate for an unsuitable measurement method. Resolve that limitation before using the index for a consequential acceptance decision.
Interpret requirements in context
Capability expectations depend on customer agreements, characteristic consequence and the applicable method. Avoid applying a familiar threshold as a universal definition of acceptable performance.
Distribution assumptions, sample uncertainty and one-sided requirements can change the analysis. Use qualified statistical support where the decision requires it.
Do not translate an index into an exact predicted defect rate without checking the assumptions supporting that interpretation.
Connect the analysis with action
Decide whether the process needs centering, reduced variation, better measurement or a different design. Define containment while the issue is addressed.
Track whether the improved condition persists in normal production. A successful demonstration should lead to a controlled operating process, not merely a completed approval document.
Before accepting the next capability report, ask whether the evidence describes the output you will actually receive. The index becomes valuable when the business can explain that connection.