Stuck at D4 in 8D? —— A Five-Step Practical Method for Root Cause Locking and Measure Verification

1. Customer Return Report, Comments in One Line

Zhou, the quality manager of an automotive parts company, has been tormented by the same 8D report for two weeks.

The customer complained that a batch of steering housing blanks had oversized internal holes, causing assembly issues. Zhou's team was very cooperative: D1 problem description took two pages, D2 team list included six departments, and D3 containment involved isolating work-in-progress and conducting a full inspection of inventory overnight. However, when the report was submitted, the customer returned it two days later with a single-line comment: "Root cause analysis is not deep enough, and corrective actions lack verification data."

Zhou felt wronged: they had asked five layers of 5Why and drawn a large fishbone diagram, so how could it still be insufficient?

This is not an isolated incident. In 70% of the 8D reports returned by customers, the issues lie in the D4 to D6 stages: superficial root cause analysis, incorrect measures, and lack of verification data. D1 to D3 are about firefighting, and most teams can complete these steps. The true determinant of 8D success or failure is the "root cutting" work starting from D4. This article uses an improvement case from a die-casting company to break down D4 to D6 into five steps, explaining how to execute each step and what constitutes a satisfactory completion.

2. What Exactly Are D4 to D6 Solving?

First, let's align on the concepts. In 8D, D4, D5, and D6 correspond to:

• D4 Root Cause Analysis: Identify the true source of the problem and confirm that it is indeed the source.
• D5 Corrective Action Selection: Choose measures that can eliminate the root cause while also preventing similar issues.
• D6 Implementation and Verification: Implement the measures, use data to prove the problem is resolved, and apply the solution horizontally.

The most confusing concepts in these three stages are three pairs. The first pair: correction, corrective action, and preventive action. Correction is the elimination of identified nonconformities, such as rework or replacement; Corrective Action is the elimination of the cause of nonconformities to prevent recurrence; Preventive Action is the elimination of the potential causes of nonconformities to prevent them from happening. Many companies mistake "replacing defective products" as a corrective action in D5, which customers can easily spot.

The second pair: direct cause and root cause. A direct cause is the "physical mechanism of the problem," such as mold core wear leading to an enlarged cavity; a root cause is "why the mechanism exists," such as the lack of a mold life standard and reliance on operator judgment for mold replacement. 8D requires digging at least to the management or process level, not stopping at the equipment level.

The third pair: verification and validation. Verification is "whether the measures were implemented as planned," and validation is "whether the problem has truly been resolved." D6 requires the latter—data to back it up.

Understanding these three pairs of concepts clarifies why the customer returned Zhou's 8D: the final 5Why layer was "operator negligence," which is neither verifiable nor improvable; the measures were "strengthen inspection, strengthen training," which are detection and documentation measures that do not eliminate the root cause.

3. Five-Step Method: From Root Cause Locking to Effect Verification

Using a complete case from a die-casting company, let's break down D4 to D6 into five steps.

Step One: Return to the Problem Description, Lock Down the Analysis Boundary.

The biggest fear in root cause analysis is "aimless wandering." Start by returning to D1 and use Is/Is Not (what it is/what it is not) to nail down the problem boundary: which product, which batch, which equipment, which time period, and which failure mode. In the case, the team confirmed: it only affected batches produced by the No. 3 die-casting machine during the night shift from July 12 to 15, with the failure mode being an oversized internal hole. The other five die-casting machines had no such issues during the same period. With a clear boundary, the subsequent analysis will not deviate—focus on the No. 3 machine.

Step Two: Three Levels of Inquiry, Digging to a "Verifiable Root Cause."

First, use a fishbone diagram to cast a wide net, listing all possible causes under people, machines, materials, methods, environment, and measurement. Then, use 5Why to dig deeper into the most suspicious lines. Finally, use FTA (Fault Tree Analysis) to connect the logic and verify. The key is the level of inquiry, at least three layers:

• Phenomenon Layer: Oversized hole diameter.
• Direct Cause Layer: Mold core wear, leading to an enlarged cavity.
• Root Cause Layer: No mold life standard, mold replacement frequency relies on operator "judgment," and the No. 3 machine continued production for over 80,000 cycles without a mold change.
• System Cause Layer: Mold life is not included in the equipment preventive maintenance (PM) system, and there is no cycle counter.

The team did one thing right here: they wrote the root cause as a "statement that can be verified or falsified with data." Compare: "operator negligence" cannot be verified; "mold wear due to over 80,000 cycles without replacement" can be verified. This is the watershed between a qualified and unqualified D4.

Step Three: Three Questions for Root Cause Verification, Fail Any and It's Not the Root Cause.

The root cause must be verified, and the method is three questions:

1. Can it be reproduced? —— Reinstall the worn mold on the No. 3 machine for a trial run, and the oversized hole diameter reoccurs.
2. Is the data supportive? —— Compare two months of mold change records and hole diameter inspection data: batches exceeding 80,000 cycles have a 6.8% out-of-tolerance rate; batches within 80,000 cycles have a 0.3% out-of-tolerance rate, a significant difference.
3. Is the counter-evidence valid? —— After replacing the new mold, the first 200 pieces all had qualified hole diameters.

Only if all three questions are answered positively can the root cause be considered locked. Note that many teams fail at the second question: using "feelings" or "impressions" as evidence, which customers will not accept.

Step Four: Select Measures by Priority, Verify Each with a Correspondence Table.

The principle for selecting measures is "elimination at the source is preferred over post-event interception." The priority order is:

Priority 1 to 2 measures are about "reducing nonconformities," while Priority 4 to 5 measures are only about "screening out nonconformities." Customers want to see the former. Additionally, each root cause should have at least one corrective action (eliminating the cause of the occurred issue) and one preventive action (preventing potential causes). Use a "root cause—measure correspondence table" to verify each, ensuring no root cause is missed and no measure is unlinked.

The measure combination in the case: Correction—immediately replace the worn mold and isolate and re-inspect all work-in-progress from the No. 3 machine from July 12 to 15 (listed in D5 but separately marked as correction); Corrective Action—install a cycle counter, establish a mold life ledger, and enforce mandatory mold changes; Preventive Action—incorporate mold life into the PM plan, install cycle counters on the other five die-casting machines, and pre-purchase spare molds for similar products.

Step Five: Implementation, Verification, Horizontal Deployment, and Documentation.

Implementing measures is not just about assigning responsibilities and completion dates; it involves four critical steps:

1. Pilot Run: Run a small batch for a week to observe if the measures have any side effects, such as frequent mold changes reducing production capacity.
2. Data Verification: Compare process data for at least three months before and after the measures. In the case, the CPK of the No. 3 machine's hole diameter improved from 1.12 to 1.56, and the customer's PPM was zero for three months.
3. Horizontal Deployment: Investigate and apply the measures to similar equipment, products, and processes. In the case, cycle counters were installed on all five die-casting machines, and two other products using the same mold were also inspected.
4. Documentation: Update the control plan (add "first article inspection after mold change"), standard operating procedures (SOS), and PFMEA (include "mold life exceeded" as a failure mode) to ensure that the experience is integrated into the system, not just retained in individual minds.

4. Five Common Pitfalls, Any of Which Can Lead to a Customer Rejection

Pitfall One: Stopping 5Why after two or three layers. Stopping at "operator did not follow the procedure" is a typical case of giving up halfway. The correct approach is to continue asking: why didn't they follow the procedure? —— The procedure was unclear, the tools were inconvenient, or there was no poka-yoke. Digging to the process and management level is essential.

Pitfall Two: Stopping at "human issues." "Lack of employee responsibility" is not a root cause but an excuse. Human issues should be traced to "why the environment allows errors": are there poka-yoke devices, are the work instructions executable, and are lighting and tooling adequate? If the case had stopped at "operator forgot to change the mold," the subsequent improvement with the cycle counter would not have occurred.

Pitfall Three: Selecting only inspection and training measures. Increasing inspection only improves interception rates, not the generation of nonconformities; training only increases the probability of "doing it right this time," but errors still occur during rush orders or when operators are fatigued. Priority 1 to 2 measures are what customers want to see.

Pitfall Four: Using "no recurrence in one month" as the verification conclusion. "No recurrence" without before-and-after data comparison, sample size explanation, or exclusion of random factors has zero persuasiveness. At least provide a comparison of the same metric (CPK, PPM, nonconformity rate) before and after the improvement.

Pitfall Five: Measures list does not match the root causes. If D5 lists five measures and D4 identifies only two root causes, the extra measures are "added on the fly," exposing a lack of rigorous analysis. Use the root cause—measure correspondence table to verify each, ensuring a one-to-one match.

5. One-Line Summary

The watershed in 8D lies in D4 to D6: the root cause must withstand three questions, measures must be selected by priority, and verification must provide data comparisons. Only then can the prevention of recurrence be considered a true closed loop, and the customer will sign off on the report.

A verifiable root cause, a matched measure, and data for verification are essential for a true closed loop in 8D.

Knowledge code: 5.2.1

Version: v20260817

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.