Process Variation Too High

Cp is low even though the mean is on target — where is the variation coming from, and which source do I attack first to gain the most Cp?

Cp is below the applicable target. If the study and measurement system are valid, recentering alone cannot increase Cp; the variation must be reduced, the specification changed, or an alternate control agreed.

1. Verify SPC stability on the study window — no runs, trends, or out-of-control points.
2. Perform a Gage R&R (crossed if possible) with 3 operators × 10 parts × 3 trials.
3. Stratify the dataset into candidate streams (cavity, spindle, lane, operator, lot).
4. Run a nested variance components analysis; extract σ² contribution per source.
5. Rank sources by percent contribution — the top two typically dominate.
6. Design a screening DoE on the top process factors; hold nuisance factors constant.
7. Confirm the vital few via a follow-up factorial or response-surface study.
8. Implement the change through formal engineering change control.
9. Re-run the capability study; require ≥ 50 stable observations under the new setting.
10. Update Control Plan, PFMEA, and reaction rules with the new σ baseline.

• Pooling multiple cavities or heads into a single dataset — variance artificially inflated.
• Fixture deformation under clamp load, not visible without a load-cell check.
• Environmental cycling (temperature, humidity) inside the study window.
• Preventive maintenance overdue; wear-driven variation masked as common cause.
• Raw material supplied against a wide internal spec even if PPAP was accepted.
• Operator technique variation on manual load steps.
• Coolant concentration drift on machining lines.

• Cp is a spread metric — no amount of centering will move it. If the customer needs a higher Cp, either variation must fall or tolerance must open.
• The biggest variation source is rarely the machine itself; it is usually raw material lot-to-lot or fixture-to-fixture on multi-station lines.
• A "high variation" process that becomes acceptable when data is split by cavity was never one process — it was several processes reported as one.
• Cutting σ in half is worth twice the effort of re-centering; but variation reduction takes weeks, centering takes hours. Sequence accordingly.

• Do not attack variation before MSA is closed — you may be chasing gauge noise.
• Do not use a capability index to prioritize variation sources — use a variance-components study instead.
• Do not compare Cp before and after a specification change; the metric is not comparable.

• Cp vs Pp: Pp captures long-term variation including between-subgroup drift; a large Cp − Pp gap points to drift, not spread.
• Cp vs Cmk: Cmk is short-term machine spread; if Cmk is high but Cp is low, the loss is process-side, not machine-side.
• Cp vs Cpk: If Cp is low, no realistic Cpk is achievable regardless of centering.

1. Run MSA — eliminate gage variation first.
2. Stratify data by shift, operator, machine, lot.
3. Use a designed experiment (DoE) to attack dominant sources.
4. Use the free Process Capability Calculator to compute Cp, Cpk, Pp, Ppk, sigma level, and predicted PPM from your measurement data — no sign-up required.

• Never assign a variance-reduction project without a documented baseline σ and target σ.
• Never accept "we tightened everything" as a corrective action — require the variance components before and after.
• Never let procurement widen the material spec to "improve capability" — you are hiding the problem, not solving it.

• cavity and process-stream stratification — Cavity and stream differences can dominate variation.
• die wear and material streams — Die and material variation need process-specific investigation.
• rational subgrouping — Sampling design determines whether variation is interpretable.

• ✓Process demonstrated stable across the full study window
• ✓MSA acceptable and repeatable
• ✓Data stratified and single-stream
• ✓Variance components quantified and ranked
• ✓Top two variation sources addressed with engineering change
• ✓DoE confirmation run performed
• ✓Fresh capability study run under new baseline
• ✓Control Plan and PFMEA re-baselined

• Process condition: statistical stability is required.
• Distribution assumption: use a distribution model justified for the data.
• Confirm process stability and measurement-system adequacy before interpreting a capability index.
• Use a justified distribution model or non-normal method when the normal model is unsuitable.