Process Mean Outside Specification

The mean sits outside a spec limit and Cpk is negative — is this a measurement failure, a setup failure, or a true process failure, and what must happen in the next hour?

Cpk is negative; modeled nearest-tail PPM exceeds 500,000 under a stable normal model.

1. Trigger the reaction plan documented in the Control Plan — do not improvise.
2. Quarantine all product produced since the last confirmed in-spec check.
3. Retain a golden sample of the offending part for cross-lab confirmation.
4. Re-measure with a second gauge and a second operator; log both readings.
5. Confirm the specification against the current released drawing revision.
6. Pull the setup sheet, tool-life counter, and last preventive-maintenance record.
7. Review the last 25 SPC points for a step change, ramp, or mixture pattern.
8. Identify the shift moment; correlate with tool change, material lot, or operator handover.
9. Apply the corrective offset only after the shift mechanism is identified.
10. Repeat the capability study with ≥ 50 measurements on the corrected process.
11. Update Control Plan reaction rules and PFMEA occurrence rating.
12. Notify the customer per CSR if any suspect part shipped.

• Wrong specification revision loaded into the CMM or vision program.
• Reversed USL / LSL after a drawing change; nominal moved but spec limits did not.
• Datum shift caused by fixture wear, chip build-up, or a loose clamp.
• Tool-length offset applied twice, or applied to the wrong tool number.
• Thermal expansion during warm-up; first-off parts pass, mid-run parts fail high.
• Master / setting standard drifted between calibrations.
• Raw material hardness or thickness batch outside approved envelope.

• A negative Cpk almost never appears in isolation — an SPC violation preceded it. Find that violation before touching the machine.
• Immediate re-centering destroys the evidence needed to prevent recurrence. Capture the state first.
• Shift changes, tool changes and material-lot changes account for the majority of sudden mean excursions in machining and molding.
• If the same characteristic drifts on multiple machines simultaneously, the cause is upstream (material, method, or measurement) — not the machine.

• Do not compute Cpk on the excursion window — the process is out of control; capability indices are undefined.
• Do not use the pre-shift and post-shift data pooled together; treat them as two processes.
• Do not report a "recovered" Cpk from a sorted stream — sorting is containment, not capability.

• Cpk < 0 means the mean crossed the nearest spec limit; the sign carries information — do not report as absolute value.
• Ppk will be even worse than Cpk during an excursion because between-subgroup variation dominates.
• Cpm exposes the target-hit failure more starkly than Cpk when a nominal is specified.

1. Apply the approved reaction plan and protect the customer.
2. Verify MSA — confirm the measurement is real.
3. Re-center via offset / tooling adjustment.
4. Open an 8D.
5. 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 accept a "gauge is fine, I checked it" statement without a documented MSA repeat.
• Never re-run a capability study to overwrite the excursion data — retain the original as PPAP evidence.
• Never let production restart on the customer characteristic without a formal first-piece signoff after re-centering.

• measured geometry versus process qualification — Capability interpretation differs from qualified process evidence.

• ✓Second-gauge confirmation logged
• ✓Drawing revision and spec orientation verified
• ✓Suspect production window quarantined
• ✓Shift moment identified on the SPC chart
• ✓Root cause classified: measurement / setup / material / method / machine
• ✓Corrective offset applied only after root cause confirmed
• ✓Confirmation capability study run on fresh stable window
• ✓Control Plan reaction rules updated
• ✓PFMEA occurrence and detection re-scored
• ✓Customer notified per CSR when applicable

• Process condition: statistical stability is required.
• Distribution assumption: approximately normal data.
• PPM convention: nearest specification-limit tail.
• Sigma-shift convention: no sigma shift applied.
• Cpk alone identifies distance to the nearest specification limit; it does not determine total two-sided defects unless centering or both tail distances are known.
• Predicted PPM is a model estimate, not a replacement for observed defect data.