After a first sample is approved, buyers still need evidence that repeat production can stay within requirements. Process capability analysis compares the output of a stable process with specification limits. It is not a substitute for a sound measurement system, a representative sample or continued process control.

Confirm the process is stable first
Collect data from the intended machine, fixture, operator method and material route over a representative run. Look for trends, shifts, tool wear effects and special causes before calculating a summary index. A short hand-picked sample may hide normal variation or changeover effects. Record the conditions and time period represented by the data.
Check the measurement system
Variation in gauges, fixturing and operator technique can distort the observed process spread. Confirm that instruments are appropriate, calibrated and used consistently. For a critical dimension, ask how the part is located and whether the gauge can reproduce the drawing’s datum scheme. If measurement uncertainty is material relative to tolerance, resolve it before judging capability.
Interpret indices with context
Capability indices compare process variation with specification width, while some indices also reflect how centered the process is. Their interpretation depends on assumptions about process stability, data distribution and sampling. Ask for the sample size, time span, chart or stability evidence, calculation method and actual data—not a single unexplained number. Agree the acceptance threshold with the customer’s quality team rather than applying a universal target.
Keep monitoring after approval
Capability is a snapshot of conditions, not a lifetime guarantee. Define ongoing checks, reaction plans and notification rules for tool, machine, material or process changes. Review trends and out-of-control responses. The NIST/SEMATECH Engineering Statistics Handbook explains the relationship between process stability and capability indices; use qualified quality personnel to set the project method.
A useful supplier discussion asks what the data represent, how variation is controlled and what happens when a signal changes. That is more informative than requesting a capability certificate without context.
Further reading: NIST/SEMATECH: What is Process Capability?