Most CAPAs Close on Paper, Not on the Floor

A Tier 2 machined components supplier ran a CAPA last quarter on a bushing OD that drifted out of tolerance during a six week production run. The defect surfaced through the customer's incoming inspection, a 100 piece sort burned three days of shop labor, and the supplier quality engineer opened an 8D the same afternoon. Six weeks later the CAPA was closed. D7 had a written preventive action. D8 had three signatures. The corrective action plan referenced a revised setup sheet and a torque value update on a clamp. The customer accepted the 8D and released the next purchase order.

Three months later the same defect came back on the same part number, off the same machine, with the same supplier engineer. The customer's 8D walked right back in.

That story is not rare. It is the dominant pattern in supplier quality, and the cost of it is hidden inside line items that nobody wants to total up. A walkthrough of one CAPA from defect to auditor sign off makes the gap visible in a way that no checklist or maturity matrix does. We are going to do exactly that walkthrough live on Tuesday, May 19, at 2:00 PM Central. Forty five minutes of demo, fifteen minutes of Q&A, no slides, no fake data. Register for the live session here.

Before the demo, the three facts that determine whether a CAPA closes on paper or on the floor.

Fact One: A Paper CAPA Does Not Close Anything

The bushing CAPA above did everything an 8D form is supposed to do. D1 named a team. D2 described the problem with a quantity and a date. D3 listed the containment action. D4 identified a root cause. D5 listed a corrective action. D6 implemented it. D7 added a preventive action. D8 closed it out.

The recurrence happened anyway because the form was the deliverable, not the fix. The setup sheet revision sat in a shared drive. The torque value change did not propagate into the control plan reaction column. The PFMEA for that operation still listed the original detection rating. The operator on the second shift never saw the revised setup sheet because the training acknowledgement workflow lived in a separate tracker and the cell lead was on vacation the week the change went live.

Every one of those things would have been caught by an effectiveness verification step that actually walks the floor and confirms the change took. D7 is supposed to be that step. In practice it is usually a check box and a date, because the engineer who wrote the corrective action is the same engineer asked to verify that the corrective action worked, on the same form, the same week the customer wants the 8D back. The verification is structural. The cost of skipping it is structural too.

The pattern repeats in three places. First, in the recurrence rate of the same defect on the same part. Second, in the next audit cycle, when the auditor asks for objective evidence that the corrective action propagated into the linked documents and the engineer rebuilds the trail from email and the shared drive at 2 AM. Third, in the next customer escalation, when the customer asks why the same 8D is sitting on the desk again and the supplier has no answer that survives a second review.

The 8D vs CAPA distinction matters here. The 8D is the response format. The CAPA is the corrective and preventive action management system that has to outlive the form. When the two are conflated, the form gets closed and the system does not.

Fact Two: Tools Exist, They Do Not Talk to Each Other

The supplier in the bushing story used five tools to close the CAPA. Word for the 8D document. Excel for the 5 Why grid and the fishbone categorization. Visio for the cause and effect diagram that the customer asked for in the response. DocuSign for the D8 signatures. A separate document control system for the setup sheet revision. None of those tools knew about each other. The links between them were the engineer's memory.

The same supplier is also expected to demonstrate, during the next IATF 16949 internal audit, that the corrective action propagated into the linked control plan, the linked PFMEA, the linked process flow, and the linked work instruction. Clauses 10.2 and 8.7 require that propagation. The auditor will look for objective evidence in the form of a revision history that lines up.

Reconstructing that propagation after the fact is where the audit reconstruction cost lives. The supplier quality team in a typical 50 to 100 person manufacturing operation spends between $13,000 and $56,000 per year on the labor of rebuilding the paper trail that the tools never connected. That number is conservative. It counts the hours the supplier engineer spends pulling email threads and spreadsheet snapshots together for an upcoming audit, the hours the document control coordinator spends matching revision letters to dates, and the hours the cell lead spends explaining why the operator training record does not match the setup sheet revision date.

The dollar number that gets the customer's attention is different. A customer initiated 8D charge back for a recurring defect at a Tier 2 supplier typically lands between $40,000 and $120,000 per event. At a Tier 1 supplier with a more complex bill of materials and a higher unit price, the same charge back range moves to $250,000 to $400,000 per event. Those are the line items that turn a supplier quality issue into a finance issue.

The cost is not the tools. The cost is the integration the tools never did. The corrective action propagation is supposed to be automatic. In most quality management systems it is a procedure document with a flow chart that nobody runs.

Fact Three: The Audit Is the Receipt, Not the Goal

The framing is the part most quality teams have backwards. Auditors do not exist to find defects. They exist to verify that the quality system catches its own defects and corrects them in a way that the next audit can see. The audit is the receipt for work the quality system was already supposed to be doing.

When a CAPA closure produces an automatic audit trail, the audit becomes a half day instead of a week. When it does not, the audit becomes a forensic exercise. The difference between the two is whether the corrective action propagation, the training acknowledgement, the document revision history, and the effectiveness verification are linked by the system or reconstructed by the engineer.

The Correct module handles CAPA as a connected workflow rather than a form. 5 Why, Fishbone, and Decision Tree analyses live in the same record. The corrective action plan is linked to the documents it changes. The effectiveness verification pulls from the production data on the affected operation, not from the engineer's recollection. The audit trail is a byproduct of doing the work, not a separate task added at the end.

That changes what auditor preparation looks like. Instead of pulling email threads at 2 AM, the engineer pulls the CAPA record and the linked documents come with it. The receipt is already written.

How the Three Analysis Tools Actually Get Used

A live CAPA closeout uses 5 Why, Fishbone, and Decision Tree as different lenses on the same defect, not as substitutes for each other. The bushing OD drift defect is a useful walkthrough because it has structural causes, environmental causes, and procedural causes that each tool surfaces differently.

5 Why on the Bushing OD Drift

The 5 Why is the first lens because it forces a sequential chain. Each why answers the previous one and the chain stops when the answer crosses out of the operator's control and into the system's control.

1. Why did the bushing OD drift out of tolerance? The boring bar deflection increased over the run as the insert wore.
2. Why did the insert wear faster than the run length expected? The coolant concentration was below specification for the last third of the run.
3. Why was the coolant concentration below specification? The refractometer check at the start of each shift was logged as a value, not as a pass or fail against the control limit.
4. Why was the check logged without a control limit? The shift change sheet template did not include the upper and lower control limits for coolant concentration.
5. Why did the template not include the control limits? The control plan listed coolant concentration as a process parameter without a reaction column entry, so the form designer treated it as informational.

The fifth why points at the control plan, not the operator. That is where the corrective action lives. Adding a reaction column entry for coolant concentration, with a defined control limit and a defined action when out of range, propagates back into the shift change sheet template and into the PFMEA detection rating. None of that is visible if the 5 Why stops at the third why and blames the operator for not refilling the coolant.

Fishbone on the Same Defect

The fishbone is the second lens because it forces lateral thinking across the six classic categories. Run the bushing defect across them.

• Man: Operator did not flag the coolant reading.
• Machine: Boring bar deflection increased with insert wear.
• Method: Shift change sheet did not include control limits.
• Material: Coolant concentration drifted because of the makeup water hardness change last month.
• Measurement: Refractometer was calibrated, the reading was correct.
• Environment: Shop humidity changed with the seasonal HVAC schedule, affecting the coolant evaporation rate.

The fishbone surfaces the material cause that the 5 Why missed. The makeup water hardness change is the structural reason the coolant drifted. The fix is not just the reaction column. It is also a coolant concentration check on the day the makeup water source changes. The two analyses produce two complementary corrective actions, not competing ones.

Decision Tree on the Corrective Action Selection

The Decision Tree is the third lens because it forces the team to choose between corrective actions when there is more than one valid path. For the bushing defect, the team has three candidate corrective actions: revise the control plan, add a coolant concentration sensor with a programmable logic controller interlock, or move the operation to a different machine with closed loop coolant control.

Each branch has a cost, a lead time, an effectiveness rating, and a risk to the rest of the production schedule. The Decision Tree forces those tradeoffs into a visible structure. The team can compare the $400 control plan revision against the $18,000 sensor install against the $80,000 machine move, with effectiveness verification criteria that match the cost. The customer 8D needs an answer in a week. The Decision Tree shows which corrective action can be in place by then and which one is the longer term preventive action.

In the live demo we run all three of these analyses on one defect, with AI validation flagging the gaps as we go. The 5 Why catches the missed why if the chain bottoms out too early. The Fishbone flags an empty category if the engineer ran out of ideas in the middle. The Decision Tree refuses to close until each branch has a numeric effectiveness criterion attached.

The Document Cascade Most Teams Skip

The corrective action does not stop at the CAPA record. It cascades into linked documents and the cascade is where IATF 16949 clause 10.2.4 and AS9100 clause 10.2 grade the closure.

For the bushing defect, the cascade lives in five places.

• Control plan: The coolant concentration row gets a reaction column entry. The reaction column references the action to take when the reading is out of control limits.
• PFMEA: The detection rating for the OD diameter failure mode improves because the shift change check is now a controlled measurement. The RPN drops accordingly. The PFMEA revision history records the change.
• Process flow: The coolant concentration check is added as an in process inspection step between operations.
• Setup sheet and work instruction: The control limits are added to the shift change sheet template. The visual aid for the refractometer is updated to show the pass and fail bands.
• Training record: Every operator on the affected machine is added to a training acknowledgement workflow for the revised setup sheet. The acknowledgement is dated and signed.

When the CAPA workflow drives this cascade automatically, the engineer reviews and approves rather than rebuilds. When it does not, the cascade is a five item to do list that gets two items checked off and three left for the next audit prep cycle.

The Package module takes the cascade into PPAP territory when the corrective action triggers a re-submission. The cascade carries the same revision history into the submission package so the customer sees the same connected change set the supplier sees internally.

Effectiveness Verification That Survives the Next Run

D7 is the step where most CAPA records get audited and graded down. The verification has to be objective, not a signature.

The objective verification for the bushing defect looks like four things measured against the affected operation over a defined window.

• Recurrence rate of the failure mode: Zero occurrences across the next three production runs of the part number. The CAPA stays open until that data is collected.
• Process capability on the affected characteristic: Cpk of 1.33 or higher across the same three runs, with subgroup data pulled from the SPC log.
• Coolant concentration log compliance: Every shift change reading recorded with a pass or fail against the control limits. Zero gaps over thirty production days.
• Training acknowledgement coverage: 100 percent of operators on the affected machine signed off on the revised setup sheet within fourteen days of the revision date.

Each one is a number, not a check box. The CAPA cannot close until each one hits the threshold. When the data comes from the production system rather than from the engineer's recollection, the verification holds up at the next audit without rebuilding anything.

What the Live Demo Walks Through

The May 19 webinar is one CAPA from start to close. We open with the defect surfacing on the floor. We open the CAPA record. We run 5 Why, Fishbone, and Decision Tree with AI validation flagging the gaps. We write the corrective action. We watch the cascade propagate into the control plan, the PFMEA, the process flow, the setup sheet, and the training record. We collect the effectiveness verification data. We close the CAPA. The audit trail is already written.

No slides. No fake data. No edits. Forty five minutes of demo, fifteen minutes of Q&A. Pat Crouse presents. Register for Tuesday, May 19, at 2:00 PM Central.

If you have a CAPA that closed on paper and came back on the floor, this is the demo that shows what closing it the other way looks like.

The Correct module runs 5 Why, Fishbone, and Decision Tree analyses with AI validation, links the corrective action to the affected control plan, PFMEA, and work instructions, and produces an audit trail as a byproduct of the closure. Start free, no credit card required. Bring a CAPA that came back and we will walk it on Tuesday.

About the Author

Pat Crouse is the CTO and Co-Founder of QualityEngineer.ai. He runs the engineering org and presents the live CAPA, PPAP, and APQP demos to the supplier quality community. Connect on LinkedIn.