9.2 Statistical Process Control (SPC), Six Sigma & Lean Tools

9.2 Statistical Process Control (SPC), Six Sigma & Lean Tools

To achieve operational excellence and protect supply networks from defect propagation, supply management professionals must move beyond qualitative quality assertions and master quantitative statistical engineering tools. Statistical Process Control (SPC), Six Sigma, and Lean Manufacturing represent the integrated analytical toolkit used by world-class procurement and manufacturing organizations to minimize process variation, eliminate non-value-added waste, and assure supplier capability.

---

1. Statistical Process Control (SPC) Fundamentals

Every manufacturing and service process exhibits variation. Statistical Process Control (SPC), pioneered by Walter Shewhart at Bell Laboratories, uses statistical methods and graphical charts to monitor process performance, distinguish between types of variation, and maintain process stability.

+-----------------------------------------------------------------------------+
|                        COMMON CAUSE VS. SPECIAL CAUSE                       |
|                                                                             |
|   DIMENSION             COMMON CAUSE VARIATION      SPECIAL CAUSE VARIATION |
|   --------------------  --------------------------  ----------------------- |
|   Origin                Inherent to the system/env  Assignable, external    |
|   Predictability        Statistically stable/normal Unpredictable, erratic  |
|   Pattern               Random fluctuations around μ Non-random (runs/spikes|
|   Responsibility        Management (system redesign) Operators/Technicians  |
|   Corrective Action     Fundamental process overhaul Locate & remove cause  |
|   Examples              Ambient humidity, standard  Tool fracture, bad raw  |
|                         machine vibration, normal   material lot, operator  |
|                         operator variance           miscalibration          |
+-----------------------------------------------------------------------------+

Common Cause vs. Special Cause Variation

1. Common Cause Variation (Random / Inherent): The natural, background variation present in every stable process. It arises from countless small, uncontrollable factors (e.g., slight voltage fluctuations, normal machine clearances, subtle raw material lot variances). A process subject only to common cause variation is in a state of statistical control; its future output is predictable within statistical bounds.
2. Special Cause Variation (Assignable / Systemic): Uncontrolled, non-random disturbances caused by specific external events or system failures (e.g., a broken cutting tool, an untrained substitute operator, a batch of out-of-spec chemical resin). Special causes introduce instability and shift the process mean or expand process dispersion. When special causes are present, the process is out of statistical control.

[!CAUTION] Tampering Hazard (Deming's Funnel Experiment): Adjusting a machine or process in response to common cause variation is called tampering. Tampering actually increases total variation and destabilizes the system. Operators must only adjust processes when a special cause is statistically indicated.

Control Charts Taxonomy & Selection

A control chart plots sample statistics over time against a centerline (Process Mean mu) and two statistically calculated control limits:

• Upper Control Limit (UCL): mu + 3 * sigma_x_bar
• Lower Control Limit (LCL): mu - 3 * sigma_x_bar

+-----------------------------------------------------------------------------+
|                         CONTROL CHART SELECTION MATRIX                      |
|                                                                             |
|                             DATA TYPE                                       |
|              Variables Data                  Attributes Data                |
|       (Continuous Measurements)           (Discrete Counts / Pass-Fail)     |
|                   │                                     │                   |
|       +-----------+-----------+             +-----------+-----------+       |
|       │                       │             │                       │       |
|   Subgroup > 1          Individual (n=1)  Defectives (Units)   Defects (Counts)
|       │                       │             │                       │       |
|   • X-bar & R Chart     • I-MR Chart        • p-chart (Variable n)  • c-chart (Constant)
|     (Sample n = 2-10)     (Individual &     • np-chart (Constant n) • u-chart (Variable)
|   • X-bar & S Chart        Moving Range)                                    |
|     (Sample n > 10)                                                         |
+-----------------------------------------------------------------------------+

1. Variables Control Charts (Continuous Data: length, weight, resistance, tensile strength)

• X-bar & R Chart (Subgroup Average and Range): The standard chart for continuous variables when sample subgroup size n is small (n = 2 to 10). The X-bar chart tracks process centering/mean, while the R chart tracks process dispersion/spread.
• X-bar & S Chart (Subgroup Average and Standard Deviation): Preferred when subgroup size n > 10, using sample standard deviation s for a more accurate estimate of dispersion.
• I-MR Chart (Individual and Moving Range): Used when data is collected as single observations (n = 1), such as chemical vat batches or destructive testing.

2. Attributes Control Charts (Discrete Data: count of defective items, flaws per surface)

• p-Chart (Proportion Defective): Tracks the percentage or fraction of nonconforming units in a sample; subgroup sample size n may vary.
• np-Chart (Number Defective): Tracks the absolute count of defective units; requires a constant subgroup size n.
• c-Chart (Count of Defects): Tracks total number of defects/flaws found in a single inspection unit of fixed, constant size (e.g., number of solder voids per PCB board).
• u-Chart (Defects per Unit): Tracks the average number of defects per inspection unit when unit area or sample size varies (e.g., defects per 100 square meters of rolled steel).

Control Limits vs. Specification Limits

A fundamental concept tested on the CPSM exam is the critical difference between Control Limits and Specification Limits:

+-----------------------------------------------------------------------------+
|                   CONTROL LIMITS VS. SPECIFICATION LIMITS                   |
|                                                                             |
|   CHARACTERISTIC     CONTROL LIMITS (UCL / LCL)   SPECIFICATION LIMITS (USL / LSL)
|   -----------------  ---------------------------  ------------------------- |
|   Definition         Voice of the Process (VOP)   Voice of the Customer (VOC)
|   Determination      Calculated from empirical    Established by design     |
|                      process sample data (±3σ)    engineers / customer needs|
|   Purpose            Evaluate statistical         Evaluate part fit, form,  |
|                      stability over time          and functional usability  |
|   Location on Chart  Drawn on Control Charts      NEVER drawn on standard   |
|                                                   control charts            |
+-----------------------------------------------------------------------------+

[!IMPORTANT] Independence of Control and Specs:

• A process can be in perfect statistical control (stable and predictable) yet produce 100% defective scrap if its control limits lie entirely outside the specification limits.
• Conversely, a process can be out of statistical control (erratic and drifting) while temporarily producing parts within specification limits.

---

2. Process Capability Analysis (Cp and Cpk)

Process Capability measures the statistical ability of an in-control process to consistently produce output that conforms to engineering customer specifications (USL and LSL).

+-----------------------------------------------------------------------------+
|                        PROCESS CAPABILITY METRICS                           |
|                                                                             |
|   POTENTIAL CAPABILITY (Cp):              ACTUAL CENTERED CAPABILITY (Cpk): |
|                                                                             |
|             USL - LSL                                  USL - μ   μ - LSL    |
|       Cp = ───────────                    Cpk = min ⎛ ───────── , ──────── ⎞|
|                6σ                                   ⎝    3σ          3σ    ⎠|
|                                                                             |
|   • Measures potential width spread.      • Measures actual performance.    |
|   • Assumes process is perfectly centered.• Accounts for process mean shift.|
|   • Independent of process mean (μ).      • Cpk is always ≤ Cp.             |
+-----------------------------------------------------------------------------+

Mathematical Formulation

1. Potential Capability Index (Cp): Cp = (USL - LSL) / (6 * sigma) Where USL - LSL is the allowable specification width (Voice of Customer), and 6 * sigma is the natural process spread (Voice of Process, encompassing 99.73% of normal distribution output).

2. Actual Centered Capability Index (Cpk): Cpk = min(CPU, CPL) = min((USL - mu) / (3 * sigma), (mu - LSL) / (3 * sigma)) Where CPU is Upper Capability, CPL is Lower Capability, mu is the empirical Process Mean, and sigma is the Process Standard Deviation.

Capability Benchmark Thresholds

• Cpk < 1.00 (Incapable): The process variation width or mean shift causes the distribution tail to cross specification limits, generating continuous defective scrap.
• Cpk = 1.00 (Marginally Capable): The natural 6 * sigma process spread exactly equals the specification tolerance band (3-sigma capability, yielding 2,700 PPM defect rate if centered).
• Cpk = 1.33 (Industry Standard Minimum): Standard minimum benchmark for existing manufacturing and commercial processes (4-sigma capability, yielding <= 63 PPM defect rate).
• Cpk = 1.67 (High Capability): Required for critical aerospace, automotive safety, medical device components, and new production line qualification (5-sigma capability).
• Cpk >= 2.00 (World-Class / Six Sigma): Process spread is half the specification tolerance (6-sigma capability, yielding <= 3.4 DPMO with a 1.5-sigma drift).

+-----------------------------------------------------------------------------+
|                        CAPABILITY COMPARISON MATRIX                         |
|                                                                             |
|   SIGMA LEVEL   Cpk VALUE    DEFECT RATE (Centered)   DEFECT RATE (1.5σ Shift)
|   -----------   ---------    ----------------------   -----------------------|
|   2 Sigma       0.67         45,500 PPM               308,537 PPM            |
|   3 Sigma       1.00          2,700 PPM                66,807 PPM            |
|   4 Sigma       1.33             63 PPM                 6,210 PPM            |
|   5 Sigma       1.67            0.57 PPM                  233 PPM            |
|   6 Sigma       2.00          0.002 PPM                   3.4 PPM (DPMO)     |
+-----------------------------------------------------------------------------+

Step-by-Step Worked Numerical Calculation

Scenario: A tier-1 machining supplier produces an aerospace aluminum sleeve with an engineering specification of 75.00 mm +/- 0.30 mm.

• Upper Specification Limit (USL) = 75.30 mm
• Lower Specification Limit (LSL) = 74.70 mm
• Total Tolerance Band = 75.30 - 74.70 = 0.60 mm

A sample run of 200 parts produces:

• Process Mean (mu) = 75.10 mm (shifted higher than target midpoint of 75.00 mm)
• Process Standard Deviation (sigma) = 0.05 mm

Step 1: Calculate Potential Capability (Cp): Cp = (USL - LSL) / (6 * sigma) = (75.30 - 74.70) / (6 * 0.05) = 0.60 / 0.30 = 2.00 Interpretation: If the process were perfectly centered at 75.00 mm, it would achieve world-class 6-sigma capability (Cp = 2.00).

Step 2: Calculate Upper Capability (CPU): CPU = (USL - mu) / (3 * sigma) = (75.30 - 75.10) / (3 * 0.05) = 0.20 / 0.15 = 1.333

Step 3: Calculate Lower Capability (CPL): CPL = (mu - LSL) / (3 * sigma) = (75.10 - 74.70) / (3 * 0.05) = 0.40 / 0.15 = 2.667

Step 4: Determine Actual Capability (Cpk): Cpk = min(CPU, CPL) = min(1.333, 2.667) = 1.33

Strategic Supply Management Takeaway: Because the process mean is shifted to 75.10 mm, the process is operating closer to the USL, degrading actual capability to Cpk = 1.33. The supplier does not need to buy a new machine (since Cp = 2.00 proves the equipment is inherently precise); they simply need to adjust tooling offsets to re-center the mean back to 75.00 mm, immediately restoring Cpk to 2.00.

---

3. Six Sigma Methodology & The DMAIC / DMADV Roadmaps

Developed by Bill Smith and Bob Galvin at Motorola in 1986 and popularized by Jack Welch at General Electric, Six Sigma is a disciplined, data-driven methodology designed to eliminate defects and achieve near-perfection in manufacturing and transactional processes.

The 1.5-sigma Shift & The 3.4 DPMO Metric

Under a pure theoretical normal curve, 6-sigma corresponds to 0.002 PPM (2 defects per billion opportunities). However, empirical research demonstrated that real-world processes experience thermal, mechanical, and material drifts over time. Motorola established a standard 1.5-sigma long-term mean shift, resulting in the official Six Sigma benchmark of 3.4 Defects Per Million Opportunities (DPMO) (99.99966% yield).

+-----------------------------------------------------------------------------+
|                     THE DMAIC VS. DMADV FRAMEWORKS                          |
|                                                                             |
|   DMAIC (For Existing Process Optimization) DMADV (For New Design / DFSS)   |
|   ----------------------------------------  ------------------------------  |
|   D - Define problem & project goals        D - Define customer design goals|
|   M - Measure baseline process capability   M - Measure VOC & CTQs          |
|   A - Analyze data to find root causes      A - Analyze design alternatives |
|   I - Improve process & verify solutions    D - Design detailed product/proc|
|   C - Control process with SPC & SOPs       V - Verify design performance   |
+-----------------------------------------------------------------------------+

DMAIC Roadmap for Existing Processes

1. Define (D): Develop Project Charter, define problem statement, quantify business impact, map high-level process using SIPOC (Suppliers, Inputs, Process, Outputs, Customers), and translate Voice of the Customer (VOC) into Critical to Quality (CTQ) specifications.
2. Measure (M): Develop operational data collection plans, perform Measurement System Analysis (Gage R&R must be < 10% for acceptable systems, 10-30% marginal), and establish baseline process capability (Z-bench and Cpk).
3. Analyze (A): Perform value stream and process flow analysis, identify root causes using Ishikawa fishbone diagrams and 5 Whys, and validate statistical significance using hypothesis testing, regression analysis, and ANOVA.
4. Improve (I): Formulate creative solutions, evaluate tradeoffs via Failure Mode and Effects Analysis (FMEA), conduct Design of Experiments (DOE), pilot solutions on a test cell, and verify capability improvement.
5. Control (C): Institute Standard Operating Procedures (SOPs), implement Poka-Yoke mistake-proofing, install Statistical Process Control (SPC) tracking charts, create dynamic Control Plans, and formally transition process ownership to the operational team.

DMADV / Design for Six Sigma (DFSS)

When a process is so fundamentally broken that incremental improvement is impossible, or when engineering a brand-new product from scratch, organizations execute DMADV (Define, Measure, Analyze, Design, Verify).

---

4. The 7 Basic Quality Tools (Ishikawa's Magnificent Seven)

Kaoru Ishikawa synthesized seven foundational visual and statistical tools that enable cross-functional teams to resolve over 90% of operational quality problems:

+-----------------------------------------------------------------------------+
|                        ISHIKAWA'S 7 BASIC QUALITY TOOLS                     |
|                                                                             |
|   1. CHECK SHEET            2. PARETO CHART           3. FISHBONE DIAGRAM   |
|      Structured data           80/20 rule descending     Cause & Effect     |
|      collection at source      bar + cumulative line     (6Ms root cause)   |
|                                                                             |
|   4. HISTOGRAM              5. CONTROL CHART          6. SCATTER DIAGRAM    |
|      Frequency distribution    Time-series tracking      Correlation between|
|      showing shape/spread      against UCL/LCL bounds    Variable X and Y   |
|                                                                             |
|   7. FLOWCHART / STRATIFICATION                                             |
|      Visual process mapping & data breakdown by shift/machine/supplier      |
+-----------------------------------------------------------------------------+

1. Check Sheet: Standardized forms for real-time, manual data collection and tallying at the workstation.
2. Pareto Chart: Bar chart sorting defect categories in descending order of frequency, overlaid with a cumulative percentage line to identify the "vital few" 20% root causes generating 80% of failures.
3. Cause-and-Effect Diagram (Fishbone / Ishikawa): Brainstorming diagram organizing root causes into standard categories—the 6Ms: Manpower (Personnel), Machine (Equipment), Method (Process/SOPs), Material (Supplies), Measurement (Gages/Inspection), and Mother Nature (Environment).
4. Histogram: Bar graph plotting continuous measurement data into bins to visualize distribution central tendency, spread, skewness, and multimodal behavior.
5. Control Chart: Statistical time-series chart with centerline and 3-sigma control limits monitoring process stability.
6. Scatter Diagram: Cartesian plot examining the mathematical correlation between an independent variable X (e.g., furnace temperature) and dependent variable Y (e.g., tensile strength).
7. Flowchart / Stratification: Step-by-step graphical depiction of process sequence (Flowchart) and disaggregation of aggregate data into distinct sub-layers based on shift, machine, operator, or supplier source (Stratification).

---

5. Lean Manufacturing & Waste Elimination

Derived from the Toyota Production System (TPS) created by Taiichi Ohno and Shigeo Shingo, Lean focuses relentlessly on maximizing customer value by eliminating non-value-added waste (Muda) and optimizing process flow.

The 8 Operational Wastes (Acronym: DOWNTIME)

+-----------------------------------------------------------------------------+
|                         THE 8 WASTES OF LEAN (DOWNTIME)                     |
|                                                                             |
|   WASTE CATEGORY        DESCRIPTION & OPERATIONAL IMPACT                    |
|   --------------------  --------------------------------------------------  |
|   D - Defects           Scrap, rework, warranty repairs, and inspection     |
|   O - Overproduction    Producing more or earlier than required (WORST MUDA)|
|   W - Waiting           Idle time waiting for materials, parts, or approvals|
|   N - Non-utilized      Failing to leverage employee intelligence and ideas |
|       Talent                                                                |
|   T - Transportation    Unnecessary physical movement of goods & materials  |
|   I - Inventory         Excess raw materials, WIP, and finished stock       |
|   M - Motion            Unnecessary ergonomic movements, bending, reaching  |
|   E - Extra-processing  Performing redundant steps or exceeding customer spec
+-----------------------------------------------------------------------------+

[!IMPORTANT] Overproduction as the Fundamental Waste: Taiichi Ohno identified Overproduction as the root cause of all other wastes because producing ahead of demand consumes working capital, creates excess inventory, requires extra transportation, consumes warehouse space, and conceals underlying quality defects until massive batches are already contaminated.

Core Lean Methodologies

1. 5S Workplace Organization

5S creates a clean, visual, standardized, and safe work environment:

• Sort (Seiri): Separate necessary tools/materials from unneeded items; remove unneeded items using the Red Tag technique.
• Set in Order (Seiton): Organize necessary items systematically ("a place for everything, and everything in its place") using shadow boards and visual floor striping.
• Shine (Seiso): Clean equipment and work areas thoroughly; cleaning acts as a primary inspection mechanism to detect leaks and misalignments.
• Standardize (Seiketsu): Establish consistent visual checklists, color-coding standards, and SOPs across all workstations.
• Sustain (Shitsuke): Maintain discipline through scheduled 5S audits, employee training, and management walk-throughs.

2. Poka-Yoke (Mistake-Proofing)

Engineered mechanisms that either prevent an error from occurring (prevention) or make an error immediately obvious so the process stops before defect creation (detection). Examples include asymmetrical electrical connectors that cannot be plugged in backwards, optical interlocks preventing press closure if hands are in the die, and automated torque wrenches that log fastening completion.

3. Kanban Pull Systems

In contrast to traditional MRP "push" scheduling that manufactures based on forecasts, Kanban is a "pull" scheduling signal authorizing production or replenishment only when consumed downstream. Physical cards, electronic signals (e-Kanban), or empty two-bin containers trigger replenishment based on actual demand.

4. Kaizen & Kaizen Blitzes

Kaizen translates to continuous, incremental improvement. A Kaizen Blitz (or Kaizen Event) is a focused, 3- to 5-day cross-functional workshop where operators, engineers, and supply managers gather at the Gemba (the real place where value is created) to rapidly analyze, redesign, and implement immediate operational improvements to a targeted workstation.