Process Capability Cpk and Ppk — Don't Confuse the Two Indices

Abstract: During an audit, Ppk is asked for, but the report only contains Cpk; Cpk is 1.67, yet frequent out-of-spec conditions occur in mass production — these are all due to misinterpretations of Cpk and Ppk. Cpk describes potential capability using within-subgroup σ, while Ppk describes actual performance using overall σ. Any error in subgroup division, normality, control status, or sampling window can lead to misleading indices. This article provides a comparison of definitions, complete calculation examples, scenarios for PPAP and mass production monitoring, MSA prerequisites, and a checklist for investigating seven root causes of high Cpk but poor quality.

1. Audit Site: The Story of Two Numbers

A certain automotive parts factory submitted a PPAP: Cpk = 1.72, and the customer approved it. After three months of mass production, the Ppk for the same characteristic fluctuated between 0.95 and 1.15, leading the customer to enforce 100% inspection.

Sources of the discrepancy:

1. During the PPAP phase, the process was controlled with 25 subgroups, and the within-subgroup σ was small.
2. Mass production included shift changes, material batches, changeover startups, and restarts after material shortages — the overall σ was larger.
3. The mean drifted 0.5σ to one side, Cpk was not recalculated in real-time, while Ppk sensitively reflected the change.

Auditor's conclusion: Potential capability is acceptable, but actual performance is insufficient — the process must be stabilized before discussing capability.

2. Formulas and Meanings (for Bilateral Specifications)

Let USL, LSL, process mean μ, and standard deviation σ be defined as follows:

Cp = (USL − LSL) / (6σ) — only spread, not center Cpk = min[(USL−μ)/3σ, (μ−LSL)/3σ] — spread + center Pp / Ppk — formulas are the same as Cp/Cpk, but σ is replaced by overall standard deviation (long-term, all sources of variation)

2.1 Within-Subgroup σ vs. Overall σ

Relationship: Ppk ≤ Cpk (usually). A large gap → instability or mean drift.

2.2 Interpretation Thresholds (for reference, follow CSR)

IATF customers often require both initial and ongoing compliance — refer to the CSR.

3. Complete Calculation Example

Characteristic: Hole diameter 12.0 ± 0.10 mm (USL=12.10, LSL=11.90)

PPAP Subgroup Data (5 pieces per subgroup, 25 subgroups in total, process controlled):

• R̄ = 0.04 mm → σ̂_within = R̄/d₂ = 0.04/2.326 = 0.0172 mm
• X̄̄ = 12.02 mm

Cpk = min[(12.10−12.02)/(3×0.0172), (12.02−11.90)/(3×0.0172)] = min[0.155, 0.233] = 0.155 — wait, this indicates insufficient capability!

(For teaching: if X̄̄=12.00, then Cpk=min[0.194, 0.194]=1.94 — this demonstrates the significant impact of center position on Cpk)

Mass Production Individual Data (30 days, including 3 batches and 2 changeovers):

• s_overall = 0.028 mm, μ = 12.03 mm
• Ppk = min[(12.10−12.03)/(3×0.028), (12.03−11.90)/(3×0.028)] = min[0.83, 1.55] = 0.83

Interpretation: If the PPAP subgroup study does not represent the variations in mass production, the approval risk is shared by the OEM and the supplier — ongoing Ppk monitoring is required.

4. Prerequisites Checklist

Before calculating Cpk:

• MSA (Gage R&R) ≤10% or as per customer requirements (6.2.1)
• Control chart has no out-of-control points (at least 20-25 subgroups)
• Data is approximately normal (or use Johnson/Bootstrap methods)
• Rational subgroups — variations within the same subgroup are only due to common causes

When calculating Ppk:

• Define the time window (e.g., rolling 30 days)
• Determine if startup data should be excluded (report separately)
• Determine if stratification is needed (by shift/equipment/mold)

Reporting Cpk without meeting prerequisites — the numbers are meaningless, and can even be worse (false sense of security).

5. PPAP vs. Mass Production: Two Different Rhythms

Customer audits often check: the reasons for the gap between PPAP Cpk and mass production Ppk — be prepared to explain.

6. Seven Root Causes of High Cpk but Poor Quality

1. Specification and process mismatch — specifications are too strict, Cp<1
2. MSA failure — high data noise
3. Using Cpk for non-normal data — apply non-normal methods
4. Misusing count data — apply p/np charts
5. Inspection and production out of sync — data does not reflect the current process
6. Reporting only Cpk and ignoring Ppk — long-term drift is not exposed
7. Special causes not excluded — out-of-control points contaminate σ

Investigating each item is more effective than "tightening inspection."

7. Business Language with FPY and Ppk

For management:

Ppk dropping from 1.4 to 1.0 is equivalent to a significant increase in out-of-spec risk — if the characteristic incurs a loss of 50 yuan per piece and the daily production is 2000 pieces, the expected loss difference can be... (quantifiable)

Connect 3.3.2 FPY and 5.3.3 doing it right the first time.

8. Software and Digital Considerations

• Minitab/JMP: Distinguish between subgroup and overall capability
• MES automatically calculates Ppk: Confirm the σ algorithm and subgroup definition match the PPAP
• Kanban display: Cpk and Ppk side by side, automatically turn orange if the gap > 0.3

9. Summary

• Cpk — how capable the process is (potential)
• Ppk — how actually good the process is (real)
• Together — determine instability vs. capability insufficiency
• Before calculation — MSA, control, normality, rational subgroup
• After calculation — use Ppk to drive mass production improvements, and use Cpk to verify improvement potential

Calculating, reporting, and following up on both indices correctly — SPC transforms from "quality department arithmetic" into a common language for process management.

Knowledge code: 6.3.2

Version: v20260528

Author: Quality Think Tank