How to Be an Excellent Quality Engineer (QE)

Quality Engineer (QE) —— the builder of the quality system, the resolver of quality issues, and the driver of continual improvement.

A QE is not a "senior QC" but a professional role that solves quality issues using engineering methods.

Chapter One: Revisiting the Role of a QE

1.1 The Essential Differences Between QC and QE

Key Differentiation: QC answers "Is this product conforming or nonconforming?" QE answers "Is this process stable? Is this system complete?"

1.2 The Four Core Roles of a QE

Four Roles of a Quality Engineer (QE)

1.3 The Growth Stages of a QE

Typical Path: QE Intern → Junior QE → Independent QE → Senior QE → Senior QE / Quality Expert → Quality Director (specific titles vary by company).

Chapter Two: The Core Technical Competency System of a QE

2.1 Statistical Process Control (SPC) — The "Stethoscope" of a QE

2.1.1 Deep Understanding of Basic Concepts

SPC is not just about "drawing a few control charts." An excellent QE's understanding of SPC should reach at least the following levels:

Common Cause vs Special Cause

Common Cause — Inherent process variation (e.g., minor material fluctuations, routine environmental temperature changes)

• Characteristics: Stable, predictable, follows a normal distribution
• Responsibility: Management / System level
• Response: Improve process design (e.g., replace equipment, optimize parameters)

Special Cause — Variation caused by external factors (e.g., tool breakage, operator error, material batch issues)

• Characteristics: Unstable, unpredictable, does not follow a normal distribution
• Responsibility: On-site operation / maintenance level
• Response: Identify and eliminate specific causes

One of the Most Important Distinctions in Quality Management: Confusing common causes and special causes is the root of most quality failures. Attributing system issues to operators (treating common causes as special causes) or ignoring sudden issues (treating special causes as common causes) will not improve quality.

2.1.2 The "Eight Out-of-Control Tests" for Control Charts

Not just about "whether it exceeds the control limits"! An excellent QE must master the Western Electric Out-of-Control Tests:

Note on No. 8: Too good data may be problematic. If 15 consecutive points are within Zone C, the data may have been "selected" or the measurement system may be faulty.

2.1.3 Practical Application of Process Capability Index (Cpk)

Understanding the Meaning Behind Cpk (Experience Reference)

Traps in Practical Application:

1. Calculating Cpk directly with non-normal data is incorrect — Use Box-Cox transformation, Johnson transformation, or non-normal capability indices (e.g., Cpmk, CpQ)
2. Within-group variation vs Total variation — Cp/Cpk use within-group σ (R-bar/d2 or S-bar/c4), Pp/Ppk use total σ. If Cp is much greater than Pp, it indicates significant between-group differences (batch instability)
3. High Cpk does not mean a good process — If the control chart is out of control, the capability index is invalid
4. Sufficient sample size — At least 25 subgroups, 100+ data points

2.2 Measurement System Analysis (MSA) — The "Calibration Ruler" of a QE

A QE should not only be able to use MSA but also judge whether the measurement system meets requirements.

2.2.1 Gauge Repeatability and Reproducibility (GR&R)

Formulas (Illustrative)

• GR&R = √(EV² + AV²) — Total measurement system variation (equipment variation + operator variation)
• %GR&R = (GR&R / Total Variation) × 100%

Judgment Criteria

Common Issues in Practical Application:

What Can a QE Do?

• When %GR&R is between 10-30%, evaluate the "correction" effect of the measurement system on process capability
• When MSA is不合格, don't just "add more gauges," but first find the root cause — replace the gauge if the precision is insufficient, standardize operation methods if the operator variation is high
• The most important aspect of MSA is "acceptability," not just "reaching ≥10%"

2.2.2 Hypothesis Testing and Measurement Systems

A good QE uses hypothesis testing to analyze issues:

• Null Hypothesis H₀: No significant change in process mean before and after improvement
• Alternative Hypothesis H₁: Significant change in process mean before and after improvement
• t-test: Compare whether the means of two data sets are significantly different
• Analysis of Variance (ANOVA): Compare three or more data sets
• Chi-square Test: Compare count data (pass/fail rates, etc.)

Case: Testing the Effectiveness of Improvement Measures

• Before improvement: 50 samples, mean = 10.05 mm, standard deviation = 0.03 mm
• After improvement: 50 samples, mean = 10.01 mm, standard deviation = 0.02 mm
• Question: Is the 0.04 mm difference a real improvement or random fluctuation?
• Approach: Two-sample t-test, resulting in p-value = 0.002 (< 0.05)
• Conclusion: The improvement is significant, the mean has indeed decreased

2.3 Failure Mode and Effects Analysis (FMEA) — The "Early Warning Radar" of a QE

2.3.1 Types of FMEA

2.3.2 Core Logic of PFMEA

PFMEA Thought Process (for each process step)

1. What could go wrong? (Failure Mode)
2. What are the consequences? (Failure Effect)
3. Severity S (1～10)
4. What is the cause? (Failure Cause)
5. Occurrence O (1～10)
6. What controls are currently in place? (Prevention + Detection)
7. Detection D (1～10)
8. RPN = S × O × D (common in older versions; newer versions focus on AP, see below)
9. Rank by risk → Develop improvements for high-risk items → Re-evaluate S/O/D (and new AP requirements) after improvements

FMEA Thinking of an Excellent QE:

1. Failure modes with S severity > 8 must be prioritized regardless of RPN (safety/regulatory related)
2. Detection D is where a QE can add the most value — the inspection methods you design determine whether you can catch the failure
3. Control measures should be divided into "Prevention" and "Detection":
   • Prevention: Poka-Yoke > SPC > First Article Inspection > Patrol Inspection > Training
   • Detection: 100% online inspection > Automated inspection > Manual inspection > Statistical sampling > Visual inspection
4. FMEA is a living document — it must be updated whenever an anomaly occurs, the process changes, or new equipment is added
5. FMEA and control plans are a pair — high-risk failure modes in FMEA must have corresponding control measures in the control plan

2.3.3 Changes in FMEA under the New AIAG-VDA Standard

New version (2019 AIAG & VDA FMEA Handbook):

Practical Points: Don't get stuck on RPN numbers. The new FMEA standard removed the RPN threshold to avoid mechanical execution like "RPN=99, no action; RPN=101, immediate action."

2.4 Design of Experiments (DOE) — The "Engineering Weapon" of a QE

DOE is a hallmark capability of an excellent QE. While complex full factorial designs are not necessary, at least the following concepts should be understood.

2.4.1 When to Use DOE?

Prioritize DOE when you need to answer the following questions:

2.4.2 Overview of Common DOE Types

2.4.3 Standard DOE Process

① Define the Problem and Objectives

• Clarify Y (Response Variable): What is it? How is it measured? Is it repeatable?
• Clarify X (Factors): List all parameters that may affect Y

② Screen Factors

• Fishbone diagram + C&E matrix + expert judgment
• Quick screening with fractional factorial experiments

③ Determine the Experimental Design

• Number of factors, levels (high/low), center points
• Block division (whether to batch)
• Randomization (to eliminate unknown biases)

④ Execute the Experiment

• Follow the randomization sequence
• Strictly record environmental conditions
• Avoid unplanned changes

⑤ Analyze the Data

• Effect plot (Pareto of Effects)
• Normal probability plot (to identify significant effects)
• Analysis of Variance (ANOVA)
• Residual analysis (to validate model assumptions)

⑥ Verification and Confirmation

• Conduct confirmation experiments under optimal conditions
• Compare predicted values with actual values

DOE Golden Rule: Do not proceed to the next step until the current step is complete. Factor selection is more important than analysis tools — "Garbage in, garbage out."

Chapter Three: Engineering Methodology for Problem Solving

3.1 Deep Understanding of the 8D Problem Solving Method

8D is not just filling out forms — each stage has its own "hidden skills" for a QE.

D2 Problem Description

Common Practice: "Product has scratches" Excellent QE's Practice (IS / IS NOT Matrix):

Tool: IS/IS NOT Matrix — This is a "complete exclusion method" before DOE screening

D4 Root Cause Analysis

Three Levels of Root Causes (Example)

5 Whys Requirement for an Excellent QE:

• Each "why" must be supported by data or facts, not guesses
• At least reach the third level (system level), don't stop at the surface
• If the final "why" points to a "people" issue → continue asking "Why did the person do this? Is it due to insufficient training? Unclear SOP? Or management deficiencies?"

D5-D6 Corrective and Verification Actions

Characteristics of Permanent Corrective Actions (PCA):

1. Can eliminate the root cause (not just control the consequences)
2. Can be applied horizontally to similar products/processes
3. Have clear verification metrics and cycles
4. Mechanisms to prevent recurrence (FMEA updates, control plan updates, standardization)

Methods to Verify the Effectiveness of Actions:

• Hypothesis testing before and after improvement (p-value < 0.05 indicates effectiveness)
• Process capability comparison before and after improvement (Cpk from 0.8 to 1.33)
• Control chart monitoring (no recurrence in the following 3 months)

3.2 Classification and Escalation Mechanism for Quality Issues

An excellent QE establishes a tiered response mechanism:

3.3 Quality Cost (COQ) — The "Economic Account" of a QE

An excellent QE not only views quality issues from a technical perspective but also quantitatively analyzes them from an economic perspective.

Components of Quality Cost

Quality Cost = Prevention Cost + Appraisal Cost + Internal Failure Cost + External Failure Cost

Quality Cost 1 : 10 : 100 Rule (Magnitude Reference)

Value Proposition of a QE:

• Don't just say "Quality is important"
• Say "Adding a poka-yoke device to this process, with an investment of 2000 yuan, is expected to reduce rework losses by 8000 yuan annually — ROI=300%"
• Use the "language" of quality costs to communicate with finance/management

Chapter Four: Deep Involvement in Quality Systems

4.1 QE and ISO 9001:2015

An excellent QE is not just "implementing the system" but can understand and drive the implementation of the system.

4.2 Internal Audits — The "Health Check Doctor" of a QE

An excellent QE is often an excellent internal auditor. But the goal is not to "find problems" but to "help the system improve."

Efficient Audit Thinking for a QE:

Before the Audit

• Review the quality data of the department to be audited over the past 3 months (customer complaints, nonconformities, CAPA)
• Prepare a "risk-based audit path" (pursue risk points, not just mechanically check clauses)

During the Audit

• Ask fewer hypothetical questions (e.g., "What would you do if a nonconformity occurred?")
• Ask more verification questions (e.g., "When was the last nonconformity? Show me the handling records")
• Trust random sampling over "verbal systems" (e.g., "Can I see the inspection records from that afternoon?")
• Combine observation, interviews, and document reviews

After the Audit

• Nonconformities should have management significance, not just be a "text game" of non-compliance with clauses
• Each nonconformity must have a root cause analysis and a corrective action plan
• Follow up until closure

4.3 CAPA System (Corrective and Preventive Actions)

CAPA is the "brain" of the quality system — a system that can form a closed-loop improvement after issues are identified.

Typical CAPA Chain

1. Awareness
2. Evaluation — Severity, impact scope, urgency
3. Containment — Immediate action to stop losses (100% screening, traceability, recall, etc.)
4. Root Cause Analysis — 8D, 5 Whys, fishbone diagram, etc.
5. Corrective Action (CA) — Eliminate the root cause of the current nonconformity
6. Preventive Action (PA) — Prevent similar issues from recurring in other areas
7. Verification — Use data to prove the effectiveness of actions
8. Standardization — Update documents, horizontal deployment

Common Traps in CAPA:

• Only doing CA and not PA (treating symptoms, not the root cause)
• Too short verification period (one week's data cannot prove "it won't happen again")
• Lack of horizontal deployment (fixing one line but not the similar one)
• Closing CAPA without full closure (verification data must be seen)

Chapter Five: Practical Development of Soft Skills

5.1 Building Cross-Departmental Influence

A QE does not have direct management authority but needs to drive improvements in other departments — influence is more important than authority.

Three Levels of Influence Building:

Level One: Professional Trust

• Others' Perception: "This QE is very professional; I believe the issues he identifies."
• Building Method: Accurate data, thorough analysis, no empty talk

Level Two: Collaborative Value

• Others' Perception: "Working with this QE, issues can be resolved, and efforts won't be wasted."
• Building Method: Bring solutions to meetings, don't pass the buck, follow up consistently

Level Three: Strategic Contribution

• Others' Perception: "This QE can see issues and opportunities that we cannot."
• Building Method: Anticipate quality risks, communicate with management using quality costs

5.2 The "Seven Steps" for Driving Improvement

1. Identify the Problem — Use data to prove severity (e.g., quality loss amount)
2. Find the Root Cause — Systematic analysis using 8D, 5 Whys, FMEA, etc.
3. Develop a Plan — At least 2-3 alternative solutions, labeled with ROI
4. Gain Support — Align with key stakeholders (production supervisor, process engineer, etc.)
5. Pilot Verification — Small batch or short-term trial run, collect comparative data
6. Full Implementation — Standardize, update documents, ensure training is in