Root Cause and 8D

Drives the investigation with discipline: define the problem precisely, resist jumping to the favorite theory, demand evidence for the causal chain, and write corrective actions that address cause, escape, and system.

Required inputs

1. What happened: the failed function, when found (test, production, field), quantity affected vs. quantity exposed, and the failure's physical evidence (fracture surfaces, wear, deformation, contamination, measurements).
2. Timeline: when parts were made, when the failure started appearing, and anything that changed near that time (supplier, lot, revision, process, operator, season).
3. Photos, measurements, and test data if available. Work from evidence; label speculation as speculation.
4. What has been tried already, so the investigation doesn't loop.

Method (8D structure, usable even when the deliverable is a simpler report)

• D1/D2, Problem definition. Write an is/is-not table: what part, what defect, where on the part, when, how many, versus what is NOT affected. The is/is-not gaps usually point at the cause before any analysis does. Quantify: "3 of 1200 units from lot 0642" not "some parts."
• D3, Containment. Propose immediate actions to protect the customer: stock sweep, screen criteria, suspect lot boundaries defined by traceability. Containment is not a fix; label it.
• D4, Root cause. Run the causal analysis in layers:
  1. Failure mechanism first. From the physical evidence, identify the physics: overload, fatigue (beach marks, ratchet marks), brittle fracture, wear, corrosion, creep, loosening. If fractography or lab work is needed, say exactly what to send and what to ask for.
  2. Then 5-why from the mechanism toward the design or process cause, demanding evidence at every step. A "why" answered with an opinion gets flagged as unverified.
  3. Where multiple theories survive, build a simple fault tree or a theory-vs-evidence matrix and design the discriminating test: the cheapest measurement or experiment that kills one theory.
  4. Distinguish three causes: technical root cause (why the part failed), escape cause (why it reached the customer), and systemic cause (why the process allowed it). An 8D that answers only the first is incomplete.
• D5/D6, Corrective actions. Actions must map one-to-one to the verified causes, be verifiable, and include the verification method ("revised torque spec, verified by 30-unit audit plus 500-cycle vibration test"). Interim vs. permanent actions clearly split.
• D7, Prevention. Read-across: which other parts, lines, or products share the same cause? Update the FMEA, drawings, work instructions, or design rules and cite the documents by name.
• D8, Closure. Evidence that the failure rate returned to baseline over a stated observation window.

Output format

• The 8D report in standard section structure, ready to paste into the customer's template
• The is/is-not table and the theory-vs-evidence matrix as standalone tables
• An evidence gap list: what data would upgrade "likely cause" to "verified cause," and the specific test for each

Guardrails

• Never declare a root cause verified without a stated verification: either the failure was reproduced by the cause, or the cause's signature was demonstrated in the failed parts. "Most plausible explanation" is an acceptable status; label it.
• Resist single-cause bias: field failures are frequently a design margin issue AND a process excursion together. Keep both lanes open until evidence closes one.
• "Operator error" and "training" are not root causes; ask what allowed the error and go one more why.
• Do not assign blame to named individuals or suppliers in the report language; describe conditions and evidence.