Six Sigma Project Dying in the I Stage? —— A Case Study on Improvement Plan Screening and Piloting in an Electronics Company

1. Introduction: Smooth Sailing in the A Stage, Nearly Fatal in the I Stage

At the beginning of 2025, an electronics manufacturing company initiated a Six Sigma project: reducing the wave soldering tin penetration defect rate. This company supplies PCBA boards to home appliance clients, and over the past year, most customer complaints were about "insufficient tin penetration and virtual soldering." The customer PPM remained high, and the client issued an ultimatum—reduce the defect rate by half within six months, or the orders would be transferred elsewhere.

The project team, led by a Black Belt, performed exceptionally well in the D, M, and A stages: In the D stage, they used SIPOC to define the scope and set Y as the wave soldering tin penetration defect rate (target reduction from 1.2% to below 0.3%); in the M stage, they conducted a measurement system analysis, confirmed the consistency between visual inspection and AOI judgment, and collected data over six weeks using stratified sampling; in the A stage, they used a fishbone diagram and 5Why to narrow down the scope, and then used a two-proportion test and interaction analysis to identify the root cause—there was an interaction between the flux spray volume and preheating temperature, and the current parameter combination was on the "edge of the window." Additionally, the uneven clamping pressure of the fixtures led to insufficient tin penetration in some solder joints.

All three stages passed the review on the first attempt, and the management was confident that the project was on the verge of success. However, no one expected that the project would almost die in the I stage—the improvement stage.

After the I stage began, the project team, following their old habits, held a brainstorming session and listed eight candidate solutions on the blackboard: changing the flux brand, increasing the spray volume, raising the preheating temperature, modifying the fixtures, adding nitrogen, adjusting the chain speed, redesigning the solder pads, and adding a rework station. After a day of discussion, they "voted based on experience" and chose the fastest and cheapest solution, "raising the preheating temperature"—the experienced technicians said, "a higher temperature will ensure better tin penetration."

However, after a two-week pilot, the tin penetration defect rate not only failed to decrease but also increased from 1.2% to 1.6%, and a new defect emerged—some boards experienced "board bursting" (bubbling of the substrate). The production line was filled with complaints, and the project was criticized by management, nearly leading to its termination.

Where did the problem lie? The root cause was clearly the "interaction" and "uneven pressure," yet the solution only adjusted a single parameter and in the wrong direction, pushing the process out of the window on the other side. This case is very typical: Six Sigma projects do not fail because of the tools but because of the step from "analytical conclusions" to "improvement plans."

2. Why the I Stage Becomes a "Project Graveyard": Three Common Pitfalls

In statistical terms, the DMAIC project failure rate is highest in the I stage (Improve, improvement stage), second only to the M stage. There are three common pitfalls.

The first is a disconnect between the solution and the root cause. In the A stage, the analysis might identify "interactions between factors" or "systemic issues," but in the I stage, the team relies on old experience to adjust a single parameter. As in the case, the root cause was the interaction, but the solution focused on a single parameter, akin to shooting at a target's periphery.

The second is voting based on feelings without evaluation criteria. After brainstorming a list of solutions, the team listens to the loudest or most senior voices without quantitatively comparing the "effectiveness, cost, and risk" of the solutions. The most expensive solution is not necessarily the best, and the fastest is often the most dangerous—quick results often come from surface-level changes.

The third is skipping the pilot and rolling out the solution comprehensively. Without small-scale validation, the risk of a wrong judgment is magnified tenfold or even a hundredfold. The project team cannot afford such a loss and often has to quietly revert, leading to a significant drop in morale.

In essence, the I stage is the converter in DMAIC that transforms "knowledge" into "action." The analysis stage answers "where the problem lies," and the improvement stage answers "what solution to use, how to verify it, and how to implement it." If this step relies on guesswork, the investments in the D, M, and A stages will be wasted.

3. Case Analysis: The "Four-Step Solution Screening Method" to Revive the Project

Returning to the case. After a month of project stagnation, the quality director and the Black Belt reviewed the project and used a "four-step solution screening method" to bring it back on track. This method is not complex, but each step hits the critical points.

Step 1: Exhaustive Enumeration and Categorization of Solutions. The eight solutions were reorganized into four categories: elimination, substitution, poka-yoke, and detection, with each solution annotated to indicate "which root cause it addresses and at which stage it acts." This reorganization revealed the problem: three of the eight solutions were unrelated to the root cause, being merely "superficially useful but actually irrelevant" fillers, and were thus eliminated.

Step 2: Solution Scoring Matrix. The remaining five solutions were scored using a Pugh matrix across five dimensions: expected effectiveness, implementation cost, risk level, implementation duration, and maintainability. The weights were not determined by the project team alone but by a review panel composed of representatives from the quality, process, production, and equipment departments—effectiveness accounted for 30%, risk for 25%, cost for 20%, and duration and maintainability each for 12.5%. The scoring results showed that the top-ranked solution was the initially overlooked "optimization of flux spray volume + adjustment of fixture clamping surface," while "raising the preheating temperature" ranked second to last.

Step 3: Preliminary Risk Assessment of Solutions. A simplified FMEA was conducted for the top three solutions: listing potential new failure modes, severity, occurrence, and detection. This step completely ruled out "raising the preheating temperature"—the RPN on the high-temperature side was as high as 280, with a severity of 9 for board bursting, and the existing furnace's temperature control precision was insufficient to support a stable high-temperature window. The top-ranked combined solution's main risk was "the need to verify the consistency of clamping pressure after fixture adjustment," which was a controllable risk.

Step 4: Small-Scale Piloting and Comparative Verification. A production line was selected for piloting, and a partial factorial design was used to arrange experiments to verify the window combinations of the three key parameters: spray volume, preheating temperature, and clamping pressure. The experimental results showed that when the spray volume was increased from 7.5 to 8.2 mL/min, the preheating temperature stabilized between 105 and 110°C, and the pressure deviation after fixture adjustment was reduced from ±18% to ±5%, the tin penetration defect rate dropped to 0.21% and remained stable for four consecutive weeks. After the pilot's success, the solution was gradually rolled out to all six production lines.

In the end, the project was completed in seven months, reducing the tin penetration defect rate from 1.2% to 0.22%, and the customer PPM decreased by 63%. The financial department confirmed an annual benefit of approximately 1.8 million yuan. Compared to the initial "two-week quick fix" solution, this process seemed slower but ultimately saved the project from failure.

4. Key Points for Implementing the I Stage: From "Selecting Solutions" to "Sustaining Results"

The lessons from this case can be distilled into five key points for implementation.

First, improvement solutions must be "responsible for the root cause." Each solution should clearly state which root cause variable it changes and the expected magnitude of change. Solutions that cannot explain the mechanism should not be selected, no matter how cheap they are.

Second, the weights in the scoring matrix should be determined across departments. The quality department values effectiveness the most, the finance department values cost the most, and the production department values duration the most. Weights determined by a single department often lead to solutions being vetoed by other departments during implementation.

Third, define "what success looks like" before piloting. Before the pilot, clearly document the criteria: what defect rate reduction is considered successful, how many days of observation are needed, how the control group is set up, and who collects the data. A pilot without criteria is like a race without a finish line.

Fourth, implement solutions with a control plan. After the I stage, the project enters the C stage, and new parameters must be integrated into the control plan and standard work instructions to prevent reverting to old practices after three months. The company later included the spray volume and preheating temperature in the process parameter card and added first article inspection for each shift.

Fifth, allocate sufficient time for the I stage. Many project teams are pressured by management to "produce results quickly" in the I stage, leading to unverified solutions being rolled out, which can be even slower. The I stage typically accounts for about 30% of the entire project cycle, which is normal and not a delay.

5. One-Sentence Summary

The first three stages of DMAIC address "accurate identification," while the I stage addresses "correct implementation and stable promotion"—even the most accurate root cause identification can lead to project failure if the solution is chosen incorrectly.

A disconnect between the solution and the root cause, voting based on feelings, and skipping the pilot are the three major pitfalls in the I stage—use a scoring matrix, preliminary risk assessment, and small-scale piloting to ensure that improvement solutions are "correct and stable."

Knowledge code: 6.1.1

Version: v20260828

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