An 8D that actually closes the problem

Many 8D reports are written to get through the customer portal, not to stop the defect. The fields are filled in, the customer’s deadlines are met, the status turns to “closed”. A few months later the same complaint arrives, on the same part number or a neighbouring one. The sign that the 8D was a formality is simple: nothing in the plant looks different from before the complaint.

Below are the places where it usually breaks and what you can check at each step.

The team

An 8D written by the quality engineer alone, at a desk, already has a problem. You need people who know the process: the process engineer, someone from production on the shift where the defect appeared, maintenance if a machine is involved, logistics if stock is part of it. Check: can you say, for each member, what they actually contributed to the analysis? If a name is there only to be there, remove it.

The problem description

This is where most 8Ds are lost, because a vague description leads to a vague cause. “Nonconforming parts at the customer” is not a description. Use the “is / is not” method for what, where, when and how much:

• Which part and which defect is affected, and which similar parts are not, even though they go through the same operations.
• Where the defect was found (at the customer’s goods-in, on their assembly line, in the field) and where it was not found.
• When the first affected lot appeared, on which shift, after which change, and which earlier lots are clean.
• How many parts, what percentage of the lot, and whether the defect covers the whole part or one specific area.

The “is not” column does the heavy lifting. If the defect appears on cavity 3 and not on 1, 2 and 4, you already have half the cause. Check: could someone who was not in the meeting reproduce the defect, or at least recognise it, from the description alone, with photos and dimensions?

Containment

Containment means protecting the customer now, not solving the cause. The typical mistake is sorting the stock in your own warehouse and stopping there. The full list of places where suspect parts can sit:

• WIP on the line and between operations.
• Finished goods in the warehouse.
• Goods in transit, including the trucks that left yesterday.
• Stock at the customer and in intermediate or consignment warehouses.
• Parts sent to service or as spares, where relevant.

Define the suspect window clearly, from the last lot proven good to the first lot made after a verified temporary action. Mark the sorted parts and keep a record: how many checked, how many rejected, on what criterion. Check: does the rejection rate from sorting make sense against what the customer found? If the customer found 5 defects in one lot and your sorting found none in the neighbouring lots, either the sorting criterion is wrong or the window is.

The cause of occurrence

“Operator error” is not a root cause. It is the place where the analysis stopped. The right question is why the process allowed the operator to make the mistake: a poka-yoke is missing, the instruction is ambiguous, the fixture accepts a part loaded backwards, the cycle time leaves no room for checking. Use “5 whys” and stop only when you reach something you can change in the process, not in the person.

Verify the hypothesis before you write it down as the cause. If you can switch the defect on and off by changing a single parameter, you have a confirmed cause. If not, you have an assumption. In the report, note how you confirmed it: a test, a reproduction, a data analysis across lots.

The escape point

This is the difference between an 8D that closes the problem and one that moves it. For every defect that reached the customer there are two separate questions: why it was made, and why you did not detect it. Many reports answer only the first.

Identify the point in the flow where the defect should have been stopped (100% inspection, first-piece check, final test, visual check at packing) and analyse separately why it did not work there. An expired gauge, a check frequency too low for how the defect appears, an unclear visual criterion, a measurement system with poor repeatability. Each has its own cause and needs its own action. Sometimes there is a third question as well: why the system, meaning the FMEA and the control plan, did not foresee this failure mode.

Corrective actions

If the only corrective action is “retrain operators”, the 8D is not finished. Training helps, but its effect fades within a few weeks, especially with staff turnover. The order of preference for actions:

1. Remove the possibility of the error (change to the fixture, tool or design).
2. Automatic detection that stops the process or separates the part.
3. Visual control supported by templates, clear limits and reference parts.
4. Training, only as support for one of the actions above.

Each action has an owner, a deadline and a clear link to one of the causes, occurrence or non-detection. Check: if you delete an action from the list, which cause is left uncovered? If the answer is “none”, the action is decoration.

Verifying effectiveness

“Action implemented” does not mean “action effective”. Decide in advance how you will prove the defect no longer appears: a number of lots or a period without recurrence, test results, a process capability measured after the change. Choose the period according to how often the defect occurred. If it appeared once a month, two weeks without complaints proves nothing.

Also check whether you can still reproduce the defect with the action in place. A poka-yoke that does not stop the line when you deliberately load the wrong part is not a poka-yoke.

FMEA, control plan and read-across

This is where 8Ds get closed on paper and stay open in the system. Once the cause is confirmed, update the PFMEA: failure mode, cause, prevention and detection controls, with the ratings revised. Update the control plan and the work instructions that follow from it. If an auditor opens the PFMEA a year from now and cannot find the defect that triggered a complaint, the question will come.

Read-across means asking where else the same thing can happen: other lines with the same technology, other parts using the same fixture or tool, other plants in the group. Make a concrete list, with a yes or no for each and the action applied where needed. A table with five rows says more than the sentence “lessons learned were communicated”.

Closing

Close the 8D only when you have proof of effectiveness, the updated documents and the completed read-across list. Then recognise the team’s work, briefly and by name, because those same people will be called on for the next 8D.

If you want to compare 8D templates or approaches to the escape point, the discussion continues in the Quality & Continuous Improvement WhatsApp group.