# Quality investigation edge cases These hypothetical patterns guide questions, evidence and escalation; they establish no actual incidents, proven causes, approvals or authority to change production. Apply the root workflow and current approved QMS. Safety and legally required reporting proceed without waiting for complete RCA. ## 1. Field failure after passing inspection Treat the complaint as evidence requiring prompt investigation and regulatory assessment. Preserve returned-device custody and appropriate biohazard controls. Compare actual use conditions with test coverage and evaluate occurrence, detection and reporting biases. Three reports divided by shipments is a crude observed report rate, not the population probability of harm. Do not decide that design testing is the root cause, or that a particular malfunction is automatically MDR-reportable, before qualified assessment. ## 2. Suspect certificates of conformance Preserve original certificates, metrology evidence and statements without accusing people. Define the suspect period from evidence, not just an overdue calibration date. Validate screening-method limitations and obtain authorized independent confirmatory testing. Trace stock and shipped assemblies, involve quality/regulatory/legal owners and preserve evidence. Suspected falsification may warrant broad containment, but a screening discrepancy alone is not proof of fraud, a fixed recall duty or a mandatory supplier termination. ## 3. Stable SPC with increasing complaints Assess whether the measured characteristic and specified limits represent functional needs, then investigate both mating components, operating conditions and measurement systems. Use tolerance/clearance models with stated dependence and distribution assumptions. Do not silently recenter production or tighten specifications without change control. Specification adequacy can itself reveal a design/QMS failure warranting CAPA; it is not categorically excluded because parts meet the old drawing. ## 4. Calibration issue affecting shipped product Preserve as-found/as-left data, historical records and timestamps. Quantify bias direction, uncertainty, drift and scope using qualified metrology review; a later single calibration offset cannot automatically correct every earlier reading. Bound suspect lots and downstream exposure. Do not suppress customer notification based on invented 25%/50% tolerance-margin categories or release held units from corrected arithmetic alone. Qualified personnel decide reinspection, notifications, reporting and disposition. ## 5. Recurrence after training-based CAPA Contain the recurrence, preserve the previous record and compare failure modes and operating exposure. Reassess the causal evidence; completed training neither proves nor disproves an operator-related contribution. Investigate equipment, interfaces, work design, measurement and environment. Follow controlled reopening/linking procedures, rather than deleting the earlier root cause or hiding a failed action. Choose effective controls and monitoring covering relevant cycles; no fixed six-month period or blanket ban on training substitutes for evidence. ## 6. Disputed FAI/audit interpretation Acknowledge the observation and seek clarification of the exact current clause and contractual requirements. Provide supported disagreement through the established process without an automatic admission. Evaluate tool/source changes with the design/quality authority against the approved FAI/change-control requirements. Neither every consumable change nor no change of specification guarantees a full, partial or absent FAI obligation. ## 7. Opposing molding defects Stop unsafe experimentation and follow the authorized reaction plan. Preserve last-known approved settings and examine material, mold, equipment, measurement and cavity-specific data. Short shots plus flash suggest hypotheses, not proof of exactly three interacting causes. Any trial, teardown or designed experiment needs qualified process ownership, safety controls and an approved plan. Do not place pressure film in machinery or change pressure/temperature from this text. Verify each causal contribution and combined outcomes. ## 8. Intermittent BGA field defect Preserve returned parts and compare production versus failure-analysis methods. Validate that imaging and electrical/reliability tests detect the relevant mechanism. Voids do not universally cause every failure, and customer misuse is not excluded merely by multiple reports. Use applicable current IPC/product requirements and qualified test protocols; do not prescribe universal alloy-specific thermal cycles or void percentages. Distinguish specification thresholds, reliability acceptance and process control limits. ## 9. Sole-source safety-critical supplier Continue approved containment and engineering review while procurement evaluates development and alternatives. Supply scarcity never authorizes unsafe release. Inspection/destructive sampling needs a qualified approved plan; one forging per lot is not a universal solution. Assess actual lead times, consumption and uncertainty rather than equating a six-month buffer with three sigma. Any commercial commitment, alternative qualification or concession requires the appropriate owner. ## 10. Go/no-go records challenged during inspection Preserve the actual observation, gauge qualification and controlled inspection plan. Attribute methods can support suitable attribute SPC and can still have measurement error; they are not inherently invalid for critical characteristics. Assess uncertainty, discrimination and risk before deciding whether variable measurement is needed. Do not promise automatic conversion of every Class II/III dimension, select a chart without rational subgroups or assert an inevitable reinspection date. Prepare an accurate, timely and authorized response. ## 11. Customer rejection after a passed sample Compare the drawing, purchase terms, customer-specific requirements and methods used by both parties. A passed AQL sample does not establish all shipped parts conform, and a qualitative deburr requirement is not necessarily unenforceable. Arrange approved continuity options and define any interim acceptance criterion with authorized change control. Verify measurement and process causes; do not unilaterally adopt a vision-system threshold or assume a specific PPAP submission scope. ## 12. Cross-contamination and changed cleaning inputs Use the applicable drug/supplement regime and validated site procedures, not device QMSR by default. Hold suspect material through authorized quality action and involve qualified toxicology/regulatory personnel. A swab result in ppm is not automatically product concentration: recovery, sampled surface, equipment area, transfer, batch size and exposure matter. Do not multiply 0.8 ppm by a 15-g dose and infer a safe 12-microgram exposure or authorize release based on ADI alone. Reassess health-based limits, analytical sensitivity, change control, validation and already distributed product. ## 13. Wrong material with apparently correct certificate Quarantine suspect stock and trace installation/shipped assemblies. Match physical lot identity to certificates and assess screening/confirmatory method limits with qualified metrology personnel. Machinability observations and handheld XRF alone do not establish full alloy specification, heat treatment, mechanical properties or forgery. Preserve suspected-counterfeit evidence; do not return it, publish allegations or file database reports automatically. Engineering/customer authorities determine replacement, concession and required reporting. ## 14. CAPA backlog Triage by unresolved safety risk, product exposure, commitments and real workload. A large count can reflect under-resourcing, complexity, weak initiation criteria or ineffective actions; it does not prove over-initiation. Preserve links and audit history if records are consolidated or administratively changed, with documented approval and remaining obligations. Do not retroactively downgrade or mass-close cases to reach 40–60 records or fifteen per engineer. Management must address capacity and systemic causes; effectiveness cannot be fabricated for an audit. ## 15. Calibration shift and process validation Establish the sign, stability and uncertainty of as-found/as-left readings and the period affected. Separate indicated settings, actual physical input and validated product output. A corrected mean or Cpk above 1.33 cannot prove that OQ/PQ challenged the necessary operating envelope or that every released product was acceptable. Qualified validation/quality owners determine containment, retrospective assessment, bridging/full revalidation and documented change controls. Never edit historical measurements in place or declare no action from a single corrected operating point. ## Handoff Record facts and provenance, competing hypotheses, measurement limitations, affected/suspect/unknown quantities, containment, current deadlines, qualified decisions needed and owners. Preserve controlled records and personal/confidential data. Do not backdate, invent signatures, conceal recurrence or turn a scenario into a release or compliance certificate.