The Last Gate Before Closing 8D: How Do You Prove the Problem Is Really Gone? — Five Steps for D6 Effectiveness Verification

1. An 8D Closed for Three Months, Recurs Again

A car parts company dealt with a customer complaint last year: transmission oil seal leakage, with a batch nonconformity rate on the customer's assembly line reaching 0.8%. The 8D team spent three weeks completing the process: in D4, they used a fishbone diagram and 5Why to lock down the root cause as wear on the oil seal press fitting tooling pin. In D5, they decided on two measures: replacing the wear-resistant bushing and adding a 100% height inspection after press fitting. The report was submitted to the customer, who replied "verification passed" two weeks later. The 8D was officially closed, and everyone breathed a sigh of relief.

Three months later, the same oil seal leakage complaint reappeared at the customer's end, with a nonconformity rate of 0.5%. During the post-mortem, the quality manager asked three questions, none of which were answered: How long was the D6 verification? — "We ran a trial for two weeks, and there were no issues." How many pieces were verified? — "About three or four hundred pieces, all done during the day shift." Did you track the nonconformity rate and compare it to the pre-improvement period during the verification? — "We were just focused on the leakage, didn't specifically track it, and didn't hear of any nonconformities."

The problem lies here. Oil seal leakage is a low-frequency defect, and it just happened that none occurred in the three or four hundred pieces during the day shift over two weeks, which does not prove the measures were effective. The wear issue on the tooling bushing was only resolved for the day shift equipment; the night shift equipment did not have the bushing replaced. After three months, the wear recurred, and so did the defect.

This is not an isolated case. Many companies' 8Ds "die" at D6: the steps from D1 to D5 are carried out with great enthusiasm, but when it comes to D6 "effectiveness verification," they either gloss over it with a sentence in the report like "measures were effective after trial operation" or simply assume it passed and push the 8D to the customer. The result is that 8Ds are closed quickly, but the problems return just as fast.

2. What Exactly is D6 Verifying?

To understand D6, first look at its position in the 8D process. Among the eight steps of 8D, D4 answers "why did the problem occur," D5 answers "what measures will eliminate the root cause," D6 answers "are these measures effective," and D7 answers "how to prevent it from happening again."

Many teams mix D5 and D6 into one step: once the measures are implemented on the production line, they assume "if it's done, it's effective." These are two different things. The verification at the end of D5 confirms that the measures address the root cause—such as confirming through a small batch trial that replacing the bushing indeed eliminates wear. D6, however, verifies that the measures stably eliminate the problem under large-scale, real, and continuous production conditions, and this elimination is not due to luck or other variables coincidentally helping.

The essence of D6 is to prove three things with data: first, the problem indicators have indeed decreased; second, the decrease is due to the current measures and not other changes during the same period; third, the effect is sustainable and not just a temporary fix during the verification period.

To determine if a D6 is just going through the motions, check if the verification plan answers: what indicators are being verified, how many samples, how long the verification period is, what baseline is being compared to, and what criteria are used to determine success.

3. Five-Step Method for D6 Effectiveness Verification

Step One: Translate Each Measure into Measurable Indicators. Measures and indicators must correspond one-to-one. For example, "replacing the wear-resistant bushing" corresponds to process parameter indicators such as "wear on the tooling pin / variation in press fitting depth," and "adding 100% height inspection" corresponds to "nonconforming product interception rate / outflow nonconformity rate." The final layer is the "leakage nonconformity rate" that the customer sees. During verification, all three layers must be examined: process parameters to prove the measures are working, interception indicators to prove the defense is in place, and result indicators to prove the problem has truly disappeared. Focusing only on the final layer means missing any breaks in the earlier stages.

Step Two: Lock Down the Baseline. Verification must have a "comparison" object. The baseline is the actual level before improvement, and the data口径 must be consistent with the verification period: the same statistical scope, the same defect definition, and the same data source. Many companies use a baseline of "0.8%" from the 8D report, but this is based on the customer's statistical口径. Using the factory's internal statistical口径 for the verification period makes the two numbers incomparable, rendering any "decrease" meaningless. It's best to collect and retain baseline data specifically during the D4 stage, rather than trying to dig up old records during verification.

Step Three: Design the Verification Plan—Sample Size, Duration, and Coverage Conditions. This is the most challenging step of the five. Three principles: first, the sample size must be sufficient, and the lower the defect rate, the larger the required sample. A rough estimate is that to verify a reduction in nonconformity rate from 0.8% to below 0.3%, the verification period should have at least several thousand pieces, not just a few hundred; second, the duration should cover a complete production cycle, including both day and night shifts, conditions after changing models and materials, and the worst day of the week, as defects often appear under the worst conditions; third, start with a small-scale pilot before rolling out, choosing one production line or one piece of equipment for controlled verification, rather than implementing across the entire factory, to minimize potential side effects. The verification plan should be summarized on one page, clearly stating the five elements: verification indicators, sample size and batches, start and end times, coverage conditions, and criteria for determining success. The plan should be reviewed by the team before starting to avoid discovering discrepancies halfway through and losing all the effort.

Step Four: Use Data for Comparison, Not Feelings. Compare the data from the verification period with the baseline. For result indicators, compare the point estimate and interval of the nonconformity rate: if there were 16 nonconformities out of 2000 pieces (0.8%) before improvement and 8 nonconformities out of 4000 pieces (0.2%) after improvement, and the observation volume is sufficient, then it can be said that there is a "statistically significant decrease." For process indicators, conduct a process capability analysis on key dimensions to see if the Cpk has improved from below 1.0 to above 1.33. Use control charts to monitor the process during the verification period for any anomalies or trends. Only when all three pieces of evidence are present can it be said that "data proves effectiveness."

Step Five: Provide Conclusions and Retain Evidence. If the verification passes, the conclusion should clearly state "what was verified, what the data shows, and what the basis for the judgment is," as an attachment to the 8D closure. This is also the original evidence that customers will review during audits. If the verification fails, do not force closure; revert to D4/D5: either the measures did not truly eliminate the root cause, the root cause was not fully identified, or the measures introduced new issues. Reverting is not a failure but a way to avoid leaving problems for the future. After the conclusion, another step is to share the verification evidence package with the customer who initially complained—customers want more than just a statement of "verified"; they want the verification method, data, and judgment conclusions. Proactively sharing the D6 evidence package can save a lot of back-and-forth clarification and build customer trust for the D7 horizontal deployment.

The following table contrasts "thorough verification" and "going through the motions" for D6, which can be used for self-checking during the D6 writing process:

4. Six Common Pitfalls in D6

Pitfall One: The Shorter the Verification Period, the Better, Rushing to Close. Some teams treat D6 as the last mile of the process, pushed by customers and leaders to conclude within a week. The verification period for low-frequency defects naturally needs to be longer. Instead of closing and facing customer accountability a month later, it's better to verify for an additional two weeks initially.

Pitfall Two: Only Verifying the Result Nonconformity Rate, Not Process Stability. Result indicators are lagging; the absence of nonconformities during the verification period may just be luck. Adding control charts and process capability analysis confirms that the process itself is under control, making the effect more robust.

Pitfall Three: Estimating Sample Size by Guesswork. Drawing 100 pieces before and after improvement, with zero nonconformities in both, and writing "verification passed"—this is the most typical fake verification. The lower the defect rate, the larger the required sample size, which is a mathematical rule, not a matter of diligence.

Pitfall Four: Mixing in Other Variables During Verification. Introducing new equipment, changing materials, or adjusting parameters during the same period makes it unclear whose contribution led to the improvement. The verification period should freeze other variables as much as possible. If this is not feasible, use segmented comparisons to separate the effects of different variables.

Pitfall Five: Only Verifying Whether Measures Were Implemented, Not Their Effectiveness. Installing inspection equipment and signing checklists, but not calculating changes in interception rates or process capability—this is D5's responsibility, while D6 focuses on effectiveness.

Pitfall Six: Not Documenting Verification. Verification data scattered in the foreman's phone or the inspector's scratch paper, with no complete records available when the 8D is closed. When the customer audits or the problem recurs, there is nothing to trace. The original records and charts from the verification period should be archived with the 8D report.

5. In a Nutshell

Whether an 8D can be closed does not depend on the time elapsed but on whether the data proves the problem has truly disappeared—any effort saved in D6 will eventually come back as a recurrence.

Use data to prove the problem is truly gone, only then is the 8D considered closed.

Knowledge code: 5.2.1

Version: v20260903

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