Root Cause Analysis Checklist

Quality engineer opens the non-conformance record in ETQ, MasterControl, Qualio, or your QMS of record. Capture part number, work order, lot/heat, operation, machine, shift, and operator. Link the customer complaint or internal defect that triggered this RCA.

Red-tag the suspect lot at the work center, in the stockroom, and in any downstream operations. Place a hold in the ERP/MES so the parts cannot be picked or shipped. Sweep adjacent lots produced on the same setup — defect boundaries are commonly wider than the reporting window.

Use Kepner-Tregoe is/is-not framing: what part, what defect, on what machine/shift, when first seen, when last seen, and explicitly what is NOT affected. A vague problem statement ("parts are out of spec") burns the next two days; a sharp one ("0.250 bore on PN 4471 oversize by 0.0008" on Op 30, second shift, since 03-Apr, only on Mazak #2") routes the investigation.

If suspect parts already shipped, escape notification is required — review the customer's quality agreement for SCAR/8D timing (often 24 hours for initial response, 10 days for full 8D). Aerospace and automotive customers under AS9100 / IATF 16949 require formal escape notification within contractual windows.

Send the initial customer notification per the quality agreement — typically a D1/D2/D3 of an 8D, including the contained quantity, suspect lot range, and proposed sort plan. Log the notification in the customer portal (IQS, NetSuite SRM, or the customer's specific SCAR system).

Export X-bar R or p-chart data for the affected characteristic from Minitab, InfinityQS, or the inline SPC system. Look for trends, shifts, or out-of-control signals (Western Electric / Nelson rules) leading up to the first defective lot. Include FAI records for the most recent setup.

Manufacturing engineer and quality engineer walk the operation with the operator. Watch a full cycle. Look at fixturing, tool wear, coolant, program rev, gauging method, and material lot. Conference-room RCAs miss what the floor sees in five minutes.

Before chasing a process issue, rule out a measurement issue. Confirm calibration is current on the gauge, CMM, or fixture used to detect the defect. Run a quick gauge R&R if the defect is borderline — MSA failures masquerade as process problems more often than people expect.

Pull ECNs, tooling changes, material lot changes, supplier changes, program revisions, and PM events from the CMMS and PLM (Arena, Windchill, SolidWorks PDM). Cross-reference against the first-defect timestamp. "What changed?" is the highest-yield question in manufacturing RCA.

Talk to operators on every shift that ran the job, not just the shift that found the defect. Ask what felt different — chatter, coolant smell, chip color, alarm history, fixture clamp force. Operators usually know; nobody asked.

Run an Ishikawa session with the cross-functional team across Man, Machine, Method, Material, Measurement, Mother Nature. Capture every plausible cause without filtering. Photo the whiteboard or use a Miro/Lucid template — attach to the QMS record.

For each top branch on the fishbone, ask why until you reach a system cause — a missing control, an unverified procedure, a training gap. Stop at "operator error" and you'll be back in 90 days; push past it to why the process allowed the error.

Reproduce the defect by deliberately introducing the suspected cause, or eliminate it and confirm the defect goes away. A theory that hasn't been turned on and off is still a theory. Document setup, parts run, and measurements.

Document the confirmed root cause(s) and the escape cause (why detection didn't catch it) separately — 8D D4 requires both. Update the PFMEA with the new failure mode, occurrence, and detection ratings; if RPN now exceeds the action threshold, flag for control-plan revision.

For each root cause, draft a corrective action that eliminates the cause and a preventive action that prevents recurrence on similar parts/processes. Prefer poka-yoke and process control changes over training-only solutions; retraining alone is the most common reason CARs reopen.

Revise the control plan with the new control method, frequency, and reaction plan. Update the operator work instruction in document control (Arena, MasterControl, Qualio) and bump the rev. If it's an AS9100 or IATF 16949 customer, control plan revisions may require customer notification under the quality agreement.

Each action gets a single named owner — not "production team." Set due dates based on action complexity: training within 7 days, fixture changes within 30, tooling changes per lead time. Track in the CAPA module so escalation fires when actions slip.

Quality manager, manufacturing engineer, production supervisor, and (where applicable) the supplier quality engineer review the plan. Customer-driven SCARs typically require the customer's SQE to approve the 8D D5 actions before D6 implementation.

Roll out the actions per the plan: install poka-yoke, change fixture, update CNC program, requalify supplier. Confirm each action is in place before moving to verification — it is common for an action to be marked "complete" in the QMS but not actually present at the machine.

Deliver training on the revised work instruction to every operator who runs the part, on every shift. Capture acknowledgment signatures in the LMS or training matrix; an ECN-driven defect that recurs because second shift wasn't trained is the canonical preventable repeat.

Track the next 30 days or 3 production lots (whichever the customer requires) at heightened inspection frequency. SPC charts back to in-control, no escapes, no related NCRs. Customer 8Ds typically require 30/60/90 day evidence of effectiveness.

Quality manager reviews the verification window evidence. If the defect rate is back to baseline and the control is holding, close the CAR. If the defect recurred or the control didn't take, reopen and escalate — closure without effectiveness is the audit finding ISO and IATF surveillance auditors most reliably catch.

If the action did not hold, restart at the fishbone — the original root cause analysis missed something. Add the case to the next management review agenda and notify the customer if the reopen affects an ongoing 8D commitment.

Identify other parts, machines, or product families that share the failure mode and apply the same control. Update standard work, FMEA library, and the lessons-learned database. The value of an RCA is in the read-across, not just the one part it fixed.