Negative Cpk: Causes and Corrective Actions

The mean is outside spec. What do I contain, verify, and correct — in that order — before recalculating?

Cpk < 0 means the process mean is outside the nearest limit. Under the normal-model assumptions, more than 50% of output is predicted beyond that limit.

The result signals a high risk of nonconforming output. Appropriate actions may include stopping shipment, containment, sorting, customer notification, and corrective action, depending on actual product risk and the governing reaction plan.

• Typical PPM: > 500,000 PPM (more than half the output)
• Sigma equivalent: < 0 (mean outside spec)

1. Trigger the customer-protection reaction plan (containment, sort, notify).
2. Freeze all suspect parts, WIP, and finished-goods in a single lot.
3. Run a same-day MSA on the gauge used for the study.
4. Confirm USL/LSL against the drawing revision on file, not against the report.
5. Segregate data by cavity, spindle, tool, and operator; recompute per stream.
6. If the mean is truly outside spec, identify the shift mechanism (tool wear, offset drift, thermal, material lot).
7. Apply the correction; run a short verification study before releasing production.
8. Update the Control Plan reaction limits so the shift is detected earlier next time.

• Calculating on the wrong spec (using nominal as a limit).
• Sigma estimated from a tiny or non-stable sample.
• Reversed USL/LSL in the report.

• Wrong nominal treated as a spec limit.
• Gauge zeroed on a drifted master; every reading is biased.
• Cavity or spindle mis-labeled — one lane is far off center, dragging the pooled mean.
• Post-plate or post-heat-treat dimension analyzed against pre-treatment spec.
• Data captured during setup or purge; the process was never at steady state.

• Negative Cpk is almost never a "capability" problem — it is a containment and root-cause event.
• A single reversed LSL/USL entry in the spec column reliably produces a large negative Cpk. Check the report before believing the process is dead.
• Thermal expansion on the first hour of a machining shift commonly pushes the mean beyond spec until warm-up completes; the study should exclude that transient with a documented cause.

• Do not use Cpk to prioritize action here — priority is containment, not statistics.
• Do not compute Cpk on the "recovered" data without a stability check first.

• A negative Cpk usually corresponds to Cp still being acceptable; the loss is 100% centering.
• Cpk < 0 implies PPM(nearest) > 500,000 under the normal model — the process cannot be shipped without sort.

1. Apply the approved reaction plan and protect the customer.
2. Verify the measurement system before anything else.
3. Re-center the process — adjust tooling, offsets, or setup.
4. Open an 8D and root-cause the centering loss.
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 issue a corrective action based on a single Cpk value without confirming MSA and specification.
• A negative Cpk that appears on Monday morning only is a warm-up problem, not a process problem.

• containment and traceability — Negative capability needs controlled containment evidence.

• ✓Containment in place before any statistical action
• ✓MSA re-verified after the negative-Cpk trigger
• ✓Specification cross-checked against the current drawing revision
• ✓Streams (cavities/spindles/operators) separated in the recomputation
• ✓Root cause classified as spec, MSA, centering, or variation before corrective action

• 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.