7.3 Process Performance (Pp, Ppk) and Long-Term vs. Short-Term Variation

7.3 Process Performance (Pp, Ppk) and Long-Term vs. Short-Term Variation

Short-Term Potential vs. Long-Term Reality

In quality assurance, a machine may perform flawlessly over a two-hour test run under the watchful eye of a tooling specialist. However, when that same machine runs across three production shifts, operated by different technicians, utilizing raw material coils from different steel heats, and subjected to day-and-night ambient factory temperature swings, the total dispersion of the parts inevitably expands.

This reality creates the need for two distinct measurement methodologies:

• Short-Term Process Capability ($C_p, C_{pk}$): Evaluates what the process is capable of doing under tightly controlled, short-term conditions.
• Long-Term Process Performance ($P_p, P_{pk}$): Evaluates how the process actually performed over extended operational periods across all sources of real-world variation.

Understanding the mathematical, operational, and customer requirements governing $C_{pk}$ versus $P_{pk}$ is a cornerstone of the ASQ Certified Quality Technician body of knowledge and a primary requirement for Automotive Industry Action Group (AIAG) and Production Part Approval Process (PPAP) submissions.

---

Defining Process Performance Indices ($P_p$ and $P_{pk}$)

Process performance indices measure the relationship between the customer's specification limits and the total overall dispersion of all parts produced, without regard to subgrouping or strict statistical control.

Mathematical Formulas for $P_p$ and $P_{pk}$

$P_p = \frac{USL - LSL}{6s_{overall}}$ $P_{pu} = \frac{USL - \bar{x}}{3s_{overall}}$ $P_{pl} = \frac{\bar{x} - LSL}{3s_{overall}}$ $P_{pk} = \min(P_{pu}, P_{pl})$

Where:

• $\bar{x}$: The grand arithmetic mean of all individual observations pooled together.
• $s_{overall}$: The standard sample standard deviation of all individual observations pooled together, calculated using the standard statistical formula: $s_{overall} = \sqrt{\frac{\sum_{i=1}^N (x_i - \bar{x})^2}{N - 1}}$ (where $N$ is the total number of individual parts across all subgroups, e.g., $N = 25 \times 5 = 125$).

---

Within-Subgroup Variation vs. Overall Total Variation

The mathematical difference between capability ($C_p, C_{pk}$) and performance ($P_p, P_{pk}$) resides entirely in the standard deviation term in the denominator.

Within-Subgroup Variation ($\hat{\sigma}_{within}$)

Within-subgroup variation is estimated using $\bar{R}/d_2$ or $\bar{s}/c_4$ from rational subgroups.

• Because a rational subgroup consists of 4 or 5 consecutive parts produced over a short window (e.g., 2 minutes), tool wear is near zero, material chemistry is identical, temperature is stable, and operator technique is constant.
• Therefore, $\hat{\sigma}_{within}$ isolates pure machine repeatability (short-term common cause variation).
• It represents the "entitlement" or technical potential of the equipment.

Overall Total Variation ($s_{overall}$)

Overall variation is calculated by pooling all $N$ data points into one single dataset, ignoring subgroups.

• It captures the total variation experienced by the customer over hours, days, weeks, or production lots.
• It is mathematically decomposed into two distinct components: $\sigma_{total}^2 = \sigma_{within}^2 + \sigma_{between}^2$ Where $\sigma_{within}^2$ is the short-term within-subgroup variance, and $\sigma_{between}^2$ is the between-subgroup variance representing process drift, shift differences, and environmental changes.

Why $s_{overall}$ is Typically Greater than $\hat{\sigma}_{within}$

Because real-world manufacturing environments always experience at least minor between-subgroup drift ($\sigma_{between}^2 > 0$): $s_{overall} \ge \hat{\sigma}_{within}$

Consequently, for any operational production process: $P_p \le C_p \quad \text{and} \quad P_{pk} \le C_{pk}$

+-------------------------------------------------------------------------+
|                 VARIATION SOURCES IN MANUFACTURING                      |
|                                                                         |
|   WITHIN-SUBGROUP VARIATION (Short-Term, σ_within):                     |
|   - Instantaneous machine vibration                                     |
|   - Bearing runout                                                      |
|   - Electrical noise in sensors                                         |
|   - Short-term hydraulic pressure pulsations                            |
|                                                                         |
|   BETWEEN-SUBGROUP VARIATION (Long-Term, σ_between):                    |
|   - Cutting tool wear and insert chipping                               |
|   - Thermal growth of machine casting from morning to afternoon         |
|   - Raw material coil-to-coil hardness differences                      |
|   - Operator-to-operator setup and adjustment habits                    |
|   - Regrinding or swapping fixtures                                     |
|                                                                         |
|   TOTAL PROCESS VARIATION: s_overall = √(σ_within² + σ_between²)        |
+-------------------------------------------------------------------------+

Comprehensive Comparison: Capability vs. Performance

| Characteristic | Process Capability ($C_p, C_{pk}$) | Process Performance ($P_p, P_{pk}$) | |---|---|---|| | Time Horizon | Short-Term (instantaneous potential) | Long-Term (extended historical performance) | | Standard Deviation | Within-subgroup: $\hat{\sigma} = \bar{R}/d_2$ or $\bar{s}/c_4$ | Total pooled sample: $s = \sqrt{\frac{\sum (x_i - \bar{x})^2}{N-1}}$ | | Variation Captured | Pure common causes within subgroups | Common causes PLUS all between-subgroup shifts/drifts | | Prerequisite | Strictly requires demonstrated statistical stability | Can be computed on any dataset (even unstable) | | AIAG PPAP Application | Ongoing mass production verification | Initial process studies and pilot production runs | | Standard Target | $\ge 1.33$ for ongoing production | $\ge 1.67$ for initial pilot runs |

---

Diagnostic Ratio: Comparing $C_{pk}$ to $P_{pk}$

A quality technician can diagnose the health of a manufacturing process by examining the numerical relationship between $C_{pk}$ and $P_{pk}$.

Condition 1: $C_{pk} \approx P_{pk}$ (Indices are nearly identical, within $\sim 10%$)

• Diagnosis: Between-subgroup variation is negligible ($\sigma_{between} \approx 0$).
• Physical Reality: The process is highly stable over time. Tool wear is minimal or properly compensated, thermal drift is well-managed, setup procedures are standardized, and material batches are uniform.
• Technician Action: The process is running predictably. Maintain standard SPC monitoring.

Condition 2: $C_{pk} \gg P_{pk}$ (Short-term capability is substantially higher than performance)

• Diagnosis: Substantial between-subgroup variation exists ($\sigma_{between}^2 > 0$).
• Physical Reality: The machine has excellent inherent precision, but the process mean is wandering or drifting over time. The short-term capability ($C_{pk}$) proves the machine can produce tight tolerances, but assignable causes are inflating total long-term variation ($P_{pk}$).
• Technician Action: Do not adjust machine tolerances or rebuild the tool! Focus root-cause investigations on between-subgroup drivers:
  1. Track tool wear rates and implement automated tool wear compensation.
  2. Monitor ambient temperature and install closed-loop chiller systems for cutting fluids.
  3. Verify raw material incoming inspection certificates (hardness, melt index).
  4. Conduct operator audits to eliminate shift-to-shift setup variations.

Systematic Diagnostic Matrix

                      SHORT-TERM CAPABILITY (Cpk)

                       Low (< 1.33)             High (>= 1.33)
                 +-----------------------+-----------------------+
                 |                       |                       |
    Low (< 1.33) |    CHRONIC PROBLEM    |    PROCESS DRIFT      |
                 | Machine lacks basic   | Machine is capable,   |
L P              | precision; overhaul   | but drift/wear/setup  |
O E              | or rebuild required.  | inflates long-term s. |
N R              +-----------------------+-----------------------+
G F              |                       |                       |
- O High (>= 1.33| MATHEMATICALLY RARE   |   WORLD-CLASS SPC     |
  R              | Check calculations;   | Process is stable,    |
T M              | s_overall should not  | centered, and capable |
  A              | be less than σ_within.| across long runs.     |
                 +-----------------------+-----------------------+

---

AIAG / PPAP Industry Standards: Initial vs. Ongoing Requirements

In the automotive, aerospace, and medical device supply chains, capability requirements are formalized under standards such as the AIAG PPAP (Production Part Approval Process) Manual (4th Edition).

1. Initial Process Studies (Pilot Runs / PPAP Submissions)

Prior to full production authorization, suppliers must produce pilot components using production tooling, production speeds, production cycle times, and production personnel:

• Sample Requirement: Typically a minimum of 300 parts collected across at least 25 subgroups.
• Performance Index: Because long-term stability has not yet been demonstrated over months of production, AIAG mandates the Process Performance Index ($P_{pk}$).
• Minimum Acceptance Threshold: $P_{pk} \ge 1.67$.
• Why 1.67 instead of 1.33? An initial study represents a short sample. Setting the initial bar at 1.67 provides a statistical safety cushion for the inevitable long-term tool wear, batch variation, and drift that will occur during mass production.

2. Ongoing Production Capability

Once PPAP is approved and the production line is operating under ongoing statistical process control:

• Capability Index: Evaluated using $C_{pk}$ derived from ongoing control charts.
• Minimum Acceptance Threshold: $C_{pk} \ge 1.33$.
• Safety-Critical / Special Characteristics: Characteristics designated as safety, legal, or critical (often marked with an inverted delta $\nabla$ or diamond symbol) mandate $C_{pk} \ge 1.67$ or even $C_{pk} \ge 2.00$ throughout the entire product lifecycle.

AIAG PPAP Decision & Action Protocol

---

Step-by-Step Worked Numerical Calculations

Worked Example: Comprehensive Capability vs. Performance Analysis

Scenario: An automotive parts supplier manufactures hydraulic brake caliper bores. The blueprint specification is $45.000 \pm 0.030\text{ mm}$ ($USL = 45.030\text{ mm}, LSL = 44.970\text{ mm}$).

A PPAP runoff study collects 25 rational subgroups of $n = 5$ parts ($N = 125$ parts total). The quality technician analyzes the data and computes the following summary statistics:

• Grand Arithmetic Mean: $\bar{\bar{X}} = 45.006\text{ mm}$
• Average Subgroup Range: $\bar{R} = 0.012\text{ mm}$
• Total Sample Standard Deviation of all 125 individual parts: $s_{overall} = 0.0078\text{ mm}$
• For $n = 5$, the range unbiasing factor is $d_2 = 2.326$.

Step 1: Calculate Within-Subgroup Standard Deviation ($\hat{\sigma}_{within}$)

$\hat{\sigma}_{within} = \frac{\bar{R}}{d_2} = \frac{0.012}{2.326} = 0.005159\text{ mm}$

Step 2: Calculate Short-Term Potential Capability ($C_p$)

$\text{Tolerance Width} = USL - LSL = 45.030 - 44.970 = 0.060\text{ mm}$ $C_p = \frac{USL - LSL}{6\hat{\sigma}_{within}} = \frac{0.060}{6 \times 0.005159} = \frac{0.060}{0.030954} = 1.9383 \approx 1.94$

Step 3: Calculate Short-Term Actual Capability ($C_{pk}$)

$C_{pu} = \frac{USL - \bar{\bar{X}}}{3\hat{\sigma}_{within}} = \frac{45.030 - 45.006}{3 \times 0.005159} = \frac{0.024}{0.015477} = 1.5507 \approx 1.55$ $C_{pl} = \frac{\bar{\bar{X}} - LSL}{3\hat{\sigma}_{within}} = \frac{45.006 - 44.970}{3 \times 0.005159} = \frac{0.036}{0.015477} = 2.3260 \approx 2.33$ $C_{pk} = \min(1.55, 2.33) = 1.55$

Step 4: Calculate Long-Term Process Performance ($P_p$)

$P_p = \frac{USL - LSL}{6s_{overall}} = \frac{0.060}{6 \times 0.0078} = \frac{0.060}{0.0468} = 1.2821 \approx 1.28$

Step 5: Calculate Long-Term Process Performance Index ($P_{pk}$)

$P_{pu} = \frac{USL - \bar{x}}{3s_{overall}} = \frac{45.030 - 45.006}{3 \times 0.0078} = \frac{0.024}{0.0234} = 1.0256 \approx 1.03$ $P_{pl} = \frac{\bar{x} - LSL}{3s_{overall}} = \frac{45.006 - 44.970}{3 \times 0.0078} = \frac{0.036}{0.0234} = 1.5385 \approx 1.54$ $P_{pk} = \min(1.03, 1.54) = 1.03$

Step 6: Technical Diagnosis & PPAP Compliance Review

1. Index Summary: $C_p = 1.94$, $C_{pk} = 1.55$, $P_p = 1.28$, $P_{pk} = 1.03$.
2. PPAP Evaluation: For an initial PPAP submission, the requirement is $P_{pk} \ge 1.67$. The process achieved only $P_{pk} = 1.03$. This initial submission is REJECTED.
3. Root Cause Analysis: Notice that $C_{pk} (1.55)$ is substantially higher than $P_{pk} (1.03)$. The within-subgroup variation is tiny ($\hat{\sigma}{within} = 0.0052\text{ mm}$), but overall variation is much larger ($s{overall} = 0.0078\text{ mm}$). $\sigma_{between}^2 = s_{overall}^2 - \hat{\sigma}_{within}^2 = 0.0078^2 - 0.005159^2 = 0.00006084 - 0.00002662 = 0.00003422$ $\sigma_{between} = \sqrt{0.00003422} = 0.00585\text{ mm}$ The between-subgroup variation is actually larger than the machine's inherent repeatability! The technician reports that machine bore sizing wanders significantly across production runs (likely boring bar tool wear or spindle warm-up drift). Correcting tool wear drift will raise $P_{pk}$ to match $C_{pk}$.

---

Quality Technician Inspection Scenarios & Common Exam Traps

Real-World Shop Scenario: Plastic Injection Molding Seasonal Drift

A technician in a precision medical molding plant notices that a syringe barrel length has a short-term $C_{pk}$ of 1.72 during morning qualification. However, the 30-day performance index drops to $P_{pk} = 1.18$. The quality manager suspects mold cavity wear. The technician analyzes subgroup averages over time and discovers that the mold dimensions are completely stable, but factory humidity and cooling water temperature fluctuations between day and night shifts alter plastic shrinkage rates. Installing an automated closed-loop mold chiller stabilizes the process, eliminating between-subgroup variance and bringing $P_{pk}$ up to 1.65.

Common Exam Traps for CQT Candidates

• Exam Trap 1: Submitting $C_{pk}$ on an Initial PPAP Submission: AIAG standards strictly require $P_{pk} \ge 1.67$ for initial process studies. Reporting $C_{pk}$ on pilot runs without demonstrated stability is an automatic failure on both customer audits and ASQ exams.
• Exam Trap 2: Believing $P_p$ Requires Control Charts: $P_p$ and $P_{pk}$ describe the historical spread of past product and do not require rational subgrouping or control charts. $C_p$ and $C_{pk}$, however, strictly require proven statistical control.
• Exam Trap 3: Confusing $s_{overall}$ with $\bar{s}$: $\bar{s}$ is the average of subgroup standard deviations (used to find $\hat{\sigma}{within} = \bar{s}/c_4$). $s{overall}$ is the standard deviation of all $N$ parts combined into a single group. Do not confuse them!
• Exam Trap 4: Dividing by $N$ Instead of $N - 1$: When calculating $s_{overall}$ by hand, remember to divide the sum of squared deviations by degrees of freedom $N - 1$, not $N$.