Six Sigma DMAIC in Practice: A Comprehensive Case Study from a Defect Rate of 8.5% to 0.3%

1. Introduction: Why DMAIC?

Six Sigma has been proven to be one of the most systematic and rigorous problem-solving frameworks in global manufacturing and service industries. However, many companies face a common dilemma when implementing Six Sigma: Although the DMAIC methodology has been learned and Black Belt and Green Belt training completed, numerous details can trip up the team when executing a full project. Issues such as poorly written Charters, insufficient data, failure to identify root causes during the analysis phase, and excessively long validation periods for improvement solutions are almost inevitable in every Six Sigma project.

This article presents a real-world Six Sigma improvement project in automotive electronics, detailing the practical process of the five DMAIC stages. The case study involves a medium-sized automotive electronics supplier (referred to as Company A) that faced a significant increase in welding defect rates in 2024. The client required an 8D report within 72 hours, but the root cause was hidden in the interaction of multiple factors. The project team, led by two Black Belt candidates and supported by a cross-departmental team from process engineering, production, equipment maintenance, and quality inspection, used the Six Sigma DMAIC approach to reduce the welding defect rate from 8.5% to 0.3% over 14 weeks, achieving an annualized benefit of approximately 2.8 million yuan.

We hope that through this comprehensive project review, readers will see that DMAIC is not just a flowchart of five stages—it is a chain of interconnected decisions: the output of each stage serves as the input for the next, and any shortcuts in one stage will be paid for with increased effort in subsequent stages.

2. Company Background and Problem Description

Company A, located in East China, primarily produces Engine Control Units (ECUs) and sensor modules, with an annual production of about 1.2 million units. Its clients include several joint venture original equipment manufacturers (OEMs). In the first quarter of 2024, the company's main product—a specific model of pressure sensor—received three consecutive complaints about welding cracks from clients, causing two production line stoppages at the OEMs.

Initial statistics from Company A's quality department showed that the Through-Hole Reflow (THR) welding process defect rate for this product increased from 1.2% at the beginning of the year to 8.5% by March. The main defect modes were:

1. Cracked Solder Joints (52% of total defects): Micro-cracks appear after welding, detectable by X-Ray inspection.
2. Excessive Void Rate (31% of total defects): IPC standards require ≤25%, but some solder joints had void rates exceeding 40%.
3. Insufficient Solder Fillet Height (17% of total defects): Solder paste did not fully fill the through-holes, affecting mechanical strength.

More challenging was the fact that these three defect modes were not independent—multiple defects often coexisted in the same solder joint, and there was significant variation between production batches: the product pass rate on Monday morning could be as high as 95%, while on Wednesday night it could plummet to 82%, making it difficult to determine whether the process had changed or if the variation was due to sampling errors.

The quality director decided to initiate this project as a Six Sigma Black Belt project, led by two Black Belt candidates and supported by a cross-departmental team from process engineering, production, equipment maintenance, and quality inspection.

3. Define Stage: Locking the Problem in a Box

3.1 Problem Statement and Project Scope

At the project kick-off meeting, the team spent an entire day doing one thing: converting the vague "poor welding quality" into a measurable, time-bound, and scoped problem statement.

The final version is as follows:

In Q1 2024, the comprehensive defect rate of the pressure sensor product's THR welding process at Company A reached 8.5%, far exceeding the internal target of ≤1.0% and the client requirement of ≤0.5%. The project goal is to reduce the THR welding defect rate to ≤0.5% within 20 weeks and maintain this level for more than three months. The project scope is limited to the THR welding process on Line 3 in the SMT workshop, excluding wave soldering and selective soldering processes.

3.2 Definition of Project Y

The team defined three response variables (Y) and set their priorities:

3.3 Project Charter and Team

The Sponsor was the quality director, and the process owner was the SMT workshop supervisor. The financial representative confirmed the estimated benefits, including reduced rework costs, fewer scrap losses, and avoided downtime, with an annualized benefit of approximately 3 million yuan. The project Charter was signed off by the management review meeting, ensuring resource allocation and cross-departmental coordination.

Duration of Define Stage: 1.5 weeks. Key Outputs: Problem Statement, Project Charter, SIPOC Diagram, Preliminary Financial Benefit Estimation.

4. Measure Stage: Let the Data Speak

4.1 Data Collection Plan

The team developed a detailed Data Collection Plan, specifying the sampling frequency, sample size, responsible person, and analysis method for each data source. Considering the periodic fluctuation characteristics of the THR welding process, the team decided to collect production data continuously for two weeks, covering day and night shifts, Monday to Sunday, and three different reflow soldering ovens (A, B, and C).

4.2 Measurement System Analysis (MSA)

Before collecting process data, the team validated the measurement system. X-Ray void rate measurement used automatic measurement software, but the AI judgment threshold varied among different operators. The team organized a cross-study with three operators and 20 samples, and the results showed:

• Repeatability: The standard deviation of five repeated measurements by the same operator on the same solder joint was 0.8%, which is acceptable.
• Reproducibility: The consistency of judgments between different operators was only 82%, mainly due to the lack of a unified standard for identifying void boundaries.

The team recalibrated the AI judgment gray scale threshold, established a unified judgment work instruction, and conducted a second validation. The corrected MSA GRR was 6.7% (≤10% is acceptable). The confirmation of the measurement system is the easiest step to skip but the most critical in the entire project—if the measurement data is unreliable, all subsequent analyses are built on sand.

4.3 Process Baseline Capability Analysis

After two weeks of data collection, the team conducted a process capability analysis on Y₁ (solder joint defect rate). The data was as follows:

• Total sample size: 12,480 solder joints (from 1,560 products, 8 solder joints per product)
• Total defects: 1,061 defective solder joints
• Overall defect rate: 8.5%
• P control chart by batch showed that the process was in statistical control (no special cause variation), indicating that the 8.5% defect rate was the inherent performance of the process, not an occasional anomaly.
• Process Sigma Level: approximately 2.9σ (including a 1.5σ shift)

Compared to the Six Sigma target of 3.4 ppm (6σ), a 2.9σ level indicated a significant improvement opportunity. More importantly, the P control chart showed that while the process was stable, it was centered at 8.5%—process stability ≠ good process. This was the most important cognitive breakthrough in the measurement stage.

4.4 Stratified Analysis

The team stratified the defect data across multiple dimensions and identified several key clues:

1. Equipment Dimension: Oven C had a significantly higher defect rate (11.2%) compared to Ovens A (7.1%) and B (7.8%).
2. Time Dimension: The defect rate during the night shift (10.5%) was higher than during the day shift (6.8%), with a significant difference (p<0.01).
3. Product Dimension: There was no significant difference in defect rates between different batches of PCB boards.
4. Pad Position Dimension: Solder joints near the board edges had a higher defect rate than those in the center.

Duration of Measure Stage: 4 weeks. Key Outputs: MSA Report, Process Baseline Capability Report, Data Collection Plan, Stratified Analysis Conclusions.

5. Analyze Stage: Uncovering the True Root Causes

5.1 Cause and Effect Matrix and FMEA

Entering the Analyze stage, the team organized a cross-departmental brainstorming session and used a fishbone diagram to identify 32 potential factors from the five dimensions of people, machines, materials, methods, and environment. They then scored the correlation between each factor and the three Ys using a Cause and Effect Matrix, selecting 12 high-score factors.

For these 12 factors, the team conducted a process FMEA, calculating the Risk Priority Number (RPN) based on severity, occurrence, and detection. Factors with RPN ≥ 100 entered the statistical validation phase:

5.2 Statistical Validation

These six factors entered the statistical validation phase, where hypothesis testing and regression analysis confirmed their statistical significance.

Finding One: Systematic Temperature Deviation in Oven C

Using a two-sample t-test to compare the actual temperature values of Oven C with Ovens A and B, the team found that the actual peak temperature in Oven C was on average 12°C lower than the set value (p<0.001), and two of the four temperature zones had abnormal thermocouple feedback. Maintenance records showed that the thermocouples in Oven C had been in use for 18 months without calibration—calibration should be done every six months.

Finding Two: Nitrogen Flow is a Determining Factor for Void Rate

The team conducted a single-factor experiment on nitrogen flow in Oven A, producing 200 products at each of three levels: 15L/min, 25L/min, and 35L/min. The single-factor ANOVA results (p<0.001) showed that nitrogen flow had a highly significant impact on the void rate:

• 15L/min: Average void rate 34.2%
• 25L/min: Average void rate 18.7%
• 35L/min: Average void rate 16.1% (no significant difference from 25L/min, p=0.23)

This indicated a threshold effect of nitrogen flow—improvements plateaued beyond 25L/min, but for Oven C, the nitrogen tubing had leaks, resulting in an actual flow rate of only 60% of the set value.

Finding Three: Insufficient Solder Paste Reconditioning Time Leads to Poor Printing Consistency

Solder paste must be reconditioned to room temperature (25±3°C) before use. On-site audits revealed that operators often compressed the reconditioning time to 1 hour (the standard requires 4 hours) to meet production targets. The team compared the printing quality of solder paste reconditioned for 1 hour and 4 hours:

Insufficient reconditioning time led to higher solder paste viscosity and poor flow, resulting in inconsistent thickness after printing and increased voids during welding.

Finding Four: Correlation Between PCB Moisture Absorption and Welding Cracks

The team tracked the storage time of different batches of PCB boards from unpacking to welding. The data showed that PCB boards stored for more than 72 hours had a 3.2 times higher rate of solder joint cracks compared to those used within 48 hours. Further validation revealed that a batch of PCB boards used by Company A had already exceeded the moisture absorption limit upon arrival—although the vacuum packaging was intact, the desiccant inside had failed.

5.3 Root Cause Confirmation

Based on the above analysis, the team confirmed four major root causes (Root Cause):

1. Equipment: Thermocouple aging in Oven C caused temperature deviation, and nitrogen tubing leaks led to insufficient flow.
2. Methods: The standard for solder paste reconditioning time was not strictly enforced, and there was a lack of monitoring mechanisms.
3. Materials: Incoming PCB boards had excessive moisture absorption, and the failure of desiccants went undetected.
4. Environment: The temperature and humidity fluctuations during the night shift (28±5°C, 65±15%RH) were more severe than during the day shift, affecting welding quality.

Duration of Analyze Stage: 4 weeks. Key Outputs: Cause and Effect Matrix, FMEA, Statistical Validation Report, Root Cause Confirmation Document.

6. Improve Stage: Solution Design and Effect Validation

6.1 Improvement Solution List

Targeting the four root causes, the team developed corresponding improvement measures:

6.2 DOE Validation—Optimizing the Welding Parameter Window

After addressing the four root causes, the team had one final question: Are the current process parameters optimal? Root cause elimination could only bring the process back to a "normal" level (expected defect rate of 2～3%), but to achieve the ≤0.5% target, the process parameters needed to be finely optimized.

The team selected three key parameters from the temperature profile for a full factorial DOE (2³ full factorial design with 3 center points, totaling 11 trials):

The response variable was the comprehensive defect score (a weighted score combining void rate, solder fillet height, and solder joint appearance).

The DOE analysis revealed a significant second-order interaction effect: there was a significant interaction between peak temperature and soak time (p=0.008). Increasing the peak temperature significantly reduced defects under short soak time (60s), but the effect was weaker under long soak time (90s). The optimal parameter combination was: peak temperature 248°C, soak time 75s, and cooling slope 2.5°C/s.

6.3 Validation Batch Results

The optimized parameter combination was validated by producing 500 products on each of the three reflow soldering ovens. The results were as follows:

• Oven A: Defect rate 0.28% (original baseline 7.1%)
• Oven B: Defect rate 0.31% (original baseline 7.8%)
• Oven C: After replacing the thermocouples and repairing the nitrogen tubing, the defect rate was 0.35% (original baseline 11.2%)

The overall defect rate across the three production lines was 0.31%, far below the target of 0.5%. The process capability Cpk improved from 0.43 to 1.52, and the sigma level increased from 2.9σ to 4.5σ.

6.4 Financial Benefit Calculation

The financial representative independently calculated the project benefits:

The total project investment (including equipment modifications, DOE trials, and training) was 210,000 yuan, with a payback period of less than one month.

Duration of Improve Stage: 4 weeks. Key Outputs: Improvement Solution List, DOE Report, Validation Batch Report, Benefit Calculation Document.

7. Control Stage: Making Improvements Routine

7.1 Control Plan (Control Plan)

The team updated the control plan for the product, adding the following control items:

1. Nitrogen Flow Monitoring: Install online flow sensors, with real-time data uploaded to the MES system, and automatic alarms for deviations of ±3L/min from the set value.
2. Oven Temperature Profile Management: Establish a daily temperature measurement system, generate weekly SPC control charts, and automatically lock the equipment if the control limits are exceeded.
3. Solder Paste Reconditioning Management: Add a timer lock to the reconditioning cabinet, preventing access before 4 hours.
4. PCB Baking Standards: Add desiccant inspection to incoming inspection, and force bake for storage exceeding 48 hours (125°C/4 hours).

7.2 Process Control and Response Plan

For key process parameters (KPCs), the team set control limits and response plans:

7.3 Standardization of Documents and Training

The team updated the following documents:

• Work Instruction: Standard Work Instruction for THR Welding Process (Revised Edition)
• Equipment Inspection Form: Daily Inspection Form for Reflow Soldering Ovens (Added thermocouple calibration and nitrogen flow checks)
• Training Materials: Solder Paste Management Standards, Temperature and Humidity Control Work Instructions

A total of 42 operators and 8 process technicians were trained and assessed to ensure their skills matched the new control requirements.

7.4 Project Handover and Long-Term Tracking

The project was formally handed over to the process owner (SMT workshop supervisor). The handover content included:

• Complete project documentation package (18 documents)
• Real-time control chart dashboard address and access permissions
• Monthly audit plan for the next three months
• Lessons learned report

Three months after the project's completion, the tracking data showed that the comprehensive defect rate remained between 0.28% and 0.35%, and the process remained stable. Client complaints related to welding defects for this product were zero.

Duration of Control Stage: 2 weeks (excluding the tracking period). Key Outputs: Control Plan, Control Chart, Response Plan, Standardized Documents, Training Records.

8. Project Review and DMAIC Success Insights

8.1 Key Factors for Project Success

Reflecting on this 14-week project, the team summarized the following key success factors:

First, the Define stage clearly defined the boundaries. At the beginning of the project, the business department hoped to include wave soldering and selective soldering in the project scope, arguing that "they are all welding processes." The team insisted on focusing only on the THR process—every doubling of the scope typically quadruples the project duration. DMAIC is not about tackling everything but about continuous improvement on a clear target.

Second, the MSA was not a formality. One of the most common mistakes in the Measure stage of Six Sigma projects is skipping or simplifying the measurement system analysis. If this project had not conducted the MSA, the judgment differences among operators would have been mixed into the data as "noise," reducing the sensitivity of statistical tests and potentially missing the true root causes.

Third, the Analyze stage adhered to "statistics before action." During the project, process engineers often suggested based on experience that "adjusting the temperature profile should solve the problem." The team did not immediately adopt these suggestions but insisted on completing systematic statistical validation. It was proven that while the temperature profile was indeed one direction for optimization, the most significant root causes were equipment maintenance (thermocouples and nitrogen tubing) and material management (solder paste reconditioning and PCB moisture protection).

Fourth, the DOE in the Improve stage was not just an embellishment but a necessary step to move from good to great. After eliminating the root causes, the process defect rate had already dropped to about 2%, but it was still short of the 0.5% target. Without the DOE to finely optimize the parameter window, the team might have stopped at the 2% level.

Fifth, the Control stage is crucial for sustaining improvements. The worst outcome for a Six Sigma project is not failing to meet the target but returning to the original state after achieving the target. The most ingenious design in this project's control plan was the timer lock on the reconditioning cabinet—this is not just a document requirement but a physical poka-yoke, ensuring operators cannot take shortcuts.

8.2 Lessons Learned

Not every step went smoothly. The team also recorded several lessons:

1. Data collection took longer than expected. Initially, the team planned to collect baseline data in one week, but discovered that some data fields in the MES system were incomplete, requiring manual entry. It is recommended to complete a data availability assessment in the Define stage.
2. Resource conflicts in the Equipment Department. In the Improve stage, the Equipment Department was simultaneously working on another production line relocation project, causing a 3-day delay in thermocouple replacement. Resource coordination should be done in advance, and in case of conflicts, the project Sponsor should prioritize.
3. The importance of Operational Definition (操作性定义) was underestimated. The team's definition of "solder joint defect" in the Define stage was not precise enough, leading to some disputes in the Measure stage (e.g., whether a void rate exactly at 25% is considered a defect). It is recommended to involve the quality inspection team in the definition formulation from the early stages of the project.

8.3 Experience Replication

Company A has already replicated the successful experience of this project to two other production lines and initiated a new round of Six Sigma projects (aiming to improve the yield of wave soldering). Additionally, the company has extended the idea of the "timer lock on the reconditioning cabinet" to glue management, solder paste management, and other scenarios, upgrading control measures from reliance on "operator awareness" to "systematic poka-yoke."

The complete methodology of this project has also been included in the company's internal Six Sigma Green Belt training materials, serving as the first practical case study for new Green Belt trainees.

9. Conclusion

DMAIC is not a linear process but a scientific decision-making framework based on data and facts. From the clear definition in the Define stage, to the verification of data reliability in the Measure stage, to the confirmation of root causes in the Analyze stage, the validation of solutions in the Improve stage, and the sustained assurance in the Control stage—each stage's rigor determines the quality of work in the next stage.

The true power of Six Sigma does not lie in the depth of statistical methods but in forcing the team to solve problems in a systematic, repeatable, and disciplined manner. When a project team truly completes an entire DMAIC cycle, they gain not only data-driven improvements but also a shift in mindset—from "guessing causes based on experience" to "finding root causes with data," and from "firefighting improvements" to "preventive control."

This case from Company A is just a snapshot. Every day, countless DMAIC projects are being executed across various industries, including manufacturing, services, healthcare, and finance. Their stories may differ, but the methodology framework is the same. Mastering DMAIC means acquiring a systematic ability to break down, analyze, solve, and solidify complex problems—not just a core competency for quality engineers but a skill that any practitioner hoping to solve problems scientifically should possess.

Systematic problem-solving is the most reliable path to continuous improvement.

Knowledge code: 6.1.1

Version: v20260721

Author: Quality Think Tank Quality Think Tank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners, helping companies continuously improve their quality capabilities.