Decision Frameworks — Quality & Non-Conformance Management

Decision Frameworks — Quality & Non-Conformance Management

This reference provides the detailed decision logic, MRB processes, RCA methodology selection, CAPA lifecycle management, SPC interpretation workflows, inspection level determination, supplier quality escalation, and cost of quality calculation models for regulated manufacturing quality engineering.

All thresholds, regulatory references, and process expectations reflect quality engineering practice across FDA 21 CFR 820, IATF 16949, AS9100, and ISO 13485 environments.

1. NCR Disposition Decision Trees

1.1 Universal Disposition Flow

Every non-conformance, regardless of regulatory environment, begins with this decision sequence. The flow terminates at the first applicable disposition; do not skip levels.

START: Non-conformance identified and documented
  │
  ├─ Is the part safety-critical or regulatory-controlled?
  │   ├─ YES → Can it be reworked to FULL conformance?
  │   │         ├─ YES → REWORK with approved procedure + 100% re-inspection
  │   │         └─ NO  → SCRAP (no use-as-is permitted without formal risk assessment
  │   │                   AND regulatory/customer approval)
  │   └─ NO  → Continue
  │
  ├─ Does the non-conformance affect form, fit, or function?
  │   ├─ YES → Can it be reworked to full conformance?
  │   │         ├─ YES → Is rework cost < 60% of replacement cost?
  │   │         │         ├─ YES → REWORK
  │   │         │         └─ NO  → SCRAP (rework is not economical)
  │   │         └─ NO  → Can it be repaired to acceptable function?
  │   │                   ├─ YES → REPAIR with engineering concession + customer
  │   │                   │         approval (if required by contract/standard)
  │   │                   └─ NO  → SCRAP
  │   └─ NO  → Continue
  │
  ├─ Is the non-conformance cosmetic only?
  │   ├─ YES → Does customer spec address cosmetic requirements?
  │   │         ├─ YES → Does the part meet customer cosmetic spec?
  │   │         │         ├─ YES → USE-AS-IS with documentation
  │   │         │         └─ NO  → Customer concession required → If granted: USE-AS-IS
  │   │         │                                                → If denied: REWORK or SCRAP
  │   │         └─ NO  → USE-AS-IS with engineering sign-off
  │   └─ NO  → Continue
  │
  ├─ Is this a dimensional non-conformance within material review authority?
  │   ├─ YES → Engineering analysis: does the dimension affect assembly or performance?
  │   │         ├─ YES → REWORK or SCRAP (depending on feasibility)
  │   │         └─ NO  → USE-AS-IS with documented engineering justification
  │   └─ NO  → Continue
  │
  └─ Is this a supplier-caused non-conformance?
      ├─ YES → Is the material needed immediately for production?
      │         ├─ YES → Sort/rework at supplier's cost + USE acceptable units
      │         │         + SCAR to supplier + debit memo for sort/rework cost
      │         └─ NO  → RETURN TO VENDOR with SCAR + debit memo or replacement PO
      └─ NO  → Evaluate per the functional impact path above

1.2 FDA-Regulated Environment (21 CFR 820 / ISO 13485) Specific Logic

Medical device non-conformances carry additional requirements:

Pre-Market (Design/Development):

• Non-conformances during design verification/validation must be documented in the Design History File (DHF)
• Disposition must consider risk per ISO 14971 — severity and probability of harm to the patient
• Use-as-is is rarely acceptable for a design non-conformance; it implies the design intent is wrong
• CAPA is almost always required to prevent recurrence in production

Post-Market (Production/Field):

• Non-conformances that could affect device safety or performance require evaluation for field action (recall, correction, removal) per 21 CFR 806
• The threshold is low: if there is any reasonable possibility of harm, evaluate formally
• Document the decision NOT to file a field action as rigorously as the decision to file one
• Complaint-related non-conformances must be linked to complaint records per 820.198
• MDR (Medical Device Report) obligations: death or serious injury must be reported to FDA within 30 calendar days (5 days for events requiring remedial action)

Disposition Authority Matrix:

1.3 Automotive Environment (IATF 16949) Specific Logic

Customer Notification Requirements:

• Any non-conformance on product shipped to the customer: notification within 24 hours of discovery
• Any process change affecting fit, form, function, or performance: PPAP resubmission required
• Use-as-is disposition: typically requires a formal deviation request to the customer through their supplier portal (e.g., GM's GQTS, Ford's MQAS, Stellantis' SQP)
• Customer may accept, reject, or accept with conditions (reduced quantity, time-limited deviation)

Control Plan Integration:

• When a non-conformance reveals a gap in the control plan, the control plan must be updated as part of the corrective action
• Special characteristics (safety/significant characteristics identified with shield or diamond symbols) have zero tolerance for non-conformance: 100% containment and immediate CAPA
• The reaction plan column of the control plan specifies the predetermined response — follow it first, then investigate

Controlled Shipping Levels:

• CS-1 (Internal Controlled Shipping): Supplier adds an additional inspection/sort step beyond normal controls and submits inspection data with each shipment
• CS-2 (External Controlled Shipping): Third-party inspection at supplier's facility, at supplier's cost, with direct reporting to customer quality
• CS-1 and CS-2 are distinct from the general supplier escalation ladder — they are customer-mandated containment measures, not supplier-initiated improvements

1.4 Aerospace Environment (AS9100) Specific Logic

Customer/Authority Approval:

• Use-as-is and repair dispositions ALWAYS require customer approval per AS9100 §8.7.1
• If the customer is a prime contractor working under a government contract, the government quality representative (DCMA or equivalent) may also need to approve
• Non-conformances on parts with key characteristics require notification to the design authority
• First Article Inspection (FAI) per AS9102 becomes invalid if a non-conformance indicates the process has changed from the qualified state — partial or full FAI resubmission may be required

Counterfeit Part Prevention:

• If a non-conformance raises suspicion of counterfeit material (unexpected material composition, incorrect markings, suspect documentation), invoke the counterfeit prevention procedure per AS9100 §8.1.4
• Quarantine the suspect material in a separate area from other MRB material
• Report to GIDEP (Government-Industry Data Exchange Program) if counterfeit is confirmed
• Do not return suspect counterfeit material to the supplier — it must be quarantined and may need to be retained as evidence

Traceability Requirements:

• Aerospace non-conformances must maintain lot, batch, heat, and serial number traceability throughout the disposition process
• Scrap disposition must include documented destruction of serialized parts to prevent re-entry into the supply chain
• OASIS database updates may be required for supplier quality events

2. Root Cause Analysis Methodology Selection Guide

2.1 Selection Decision Matrix

2.2 The 5 Whys: When It Works and When It Doesn't

5 Whys works well when:

• The failure is a single event with a clear before/after state change
• Each "why" can be verified with data (measurement, observation, record review)
• The causal chain does not branch — there is a single dominant cause
• The investigation can reach a systemic cause (process, system, or design issue) within 5 iterations

5 Whys fails when:

• Multiple independent causes interact to produce the failure (combinatorial causes)
• The analyst stops at "human error" or "operator mistake" — this is never a root cause
• Each "why" is answered with opinion rather than verified data
• The analysis becomes circular (Why A? Because B. Why B? Because A.)
• Organizational pressure drives toward a "convenient" root cause that avoids systemic change

Verification protocol for each "why" level:

2.3 Ishikawa Diagram: 6M Framework Deep Dive

For each M category, specific investigation questions that separate thorough analysis from checkbox exercises:

Man (Personnel):

• Was the operator trained AND certified on this specific operation?
• When was the most recent certification renewal?
• Was this the operator's normal workstation or were they cross-trained/temporary?
• Was the shift staffing at normal levels or was this during overtime/short-staffing?
• Check operator error rate data — is this an isolated event or a pattern for this individual?

Machine (Equipment):

• When was the last preventive maintenance performed (date AND what was done)?
• Is the machine within its calibration cycle for all measuring functions?
• Were any alarms, warnings, or parameter drifts logged before the event?
• Has the machine been modified, repaired, or had a tooling change recently?
• Check the machine's historical Cpk trend — has capability been declining?

Material:

• Is this a new lot of raw material? When did the lot change?
• Were incoming inspection results within normal range, or marginal-pass?
• Does the material certificate match what was physically received (heat number, mill, composition)?
• Has the material been stored correctly (temperature, humidity, shelf life, FIFO rotation)?
• Were any material substitutions or equivalents authorized?

Method (Process):

• Is the work instruction current revision? When was it last revised?
• Does the operator actually follow the work instruction as written (observation, not assumption)?
• Were any process parameters changed recently (speeds, feeds, temperatures, pressures, cure times)?
• Was an engineering change order (ECO) recently implemented on this part or process?
• Is there a gap between the documented method and the actual method (tribal knowledge)?

Measurement:

• Was the measurement system used for this inspection validated (Gauge R&R)?
• Is the gauge within calibration? Check both certificate and physical condition.
• Was the correct measurement method used (per the control plan or inspection instruction)?
• Did the measurement environment (temperature, vibration, lighting) affect the result?
• For attribute inspections (go/no-go, visual): what is the inspection effectiveness rate?

Mother Nature (Environment):

• Were ambient conditions (temperature, humidity) within process specification?
• Were there any environmental events (power fluctuation, compressed air pressure drop, vibration from construction)?
• Is there a shift-to-shift or day-to-day correlation in the data (temperature cycling, humidity changes)?
• Was the factory HVAC system operating normally?
• For cleanroom or controlled environment processes: were environmental monitoring logs within specification?

2.4 8D Methodology: Detailed Gate Requirements

Each D-step has specific outputs required before advancing. Skipping gates creates 8Ds that look complete but don't actually solve the problem.

2.5 Fault Tree Analysis: Construction Methodology

Step 1: Define the Top Event

• State the undesired event in specific, measurable terms
• Example: "Shaft diameter exceeds USL of 25.05mm on finished machined part"
• Not: "Bad parts" or "Quality problem"

Step 2: Identify Immediate Causes (Level 1)

• What must be true for the top event to occur?
• Use AND gates (all causes must be present) and OR gates (any single cause is sufficient)
• Example: "Shaft OD too large" can be caused by (OR gate): tool wear, incorrect tool offset, material oversize, thermal expansion, fixture misalignment

Step 3: Decompose Each Cause (Levels 2–N)

• For each Level 1 cause, ask: what causes this?
• Continue decomposing until you reach basic events (events with known failure rates or that cannot be further decomposed)
• Example: "Tool wear" caused by (AND gate): extended run time + inadequate tool change interval + no in-process SPC alert

Step 4: Quantify (when data is available)

• Assign probability values to basic events using historical data, MTBF data, or engineering estimates
• Calculate top event probability through the gate logic
• Identify the minimal cut sets (smallest combinations of basic events that cause the top event)
• Focus corrective actions on the highest-probability cut sets

3. CAPA Writing and Verification Framework

3.1 CAPA Initiation Criteria

Always initiate CAPA for:

• Repeat non-conformance: same failure mode occurring 3+ times in 12 months
• Customer complaint involving product performance, safety, or regulatory compliance
• External audit finding (FDA, notified body, customer, registrar)
• Field failure or product return
• Trend signal: SPC control chart out-of-control pattern (not isolated point)
• Regulatory requirement change affecting existing products/processes
• Post-market surveillance data indicating potential safety concern

Consider CAPA (judgment call) for:

• Repeat non-conformance: same failure mode 2 times in 12 months
• Internal audit finding of moderate significance
• Supplier non-conformance with systemic indicators
• Near-miss event (non-conformance caught before reaching customer)
• Process deviation from validated parameters without product impact

Do NOT initiate CAPA for:

• Isolated non-conformance with clear, non-recurring cause (one-off tool breakage, power outage)
• Non-conformance fully addressed by NCR disposition with no systemic implication
• Customer cosmetic preference that doesn't violate any specification
• Minor documentation errors caught and corrected within the same day

3.2 CAPA Action Hierarchy (Effectiveness Ranking)

Corrective actions are not created equal. Rank by effectiveness and default to the highest feasible level:

If your corrective action is ranked 5 or 6 and a rank 1-4 action is feasible, the CAPA will likely be challenged by auditors. Training alone is never an adequate corrective action for a significant non-conformance.

3.3 CAPA Effectiveness Verification Protocol

Phase 1: Implementation Verification (within 2 weeks of target date)

Phase 2: Effectiveness Validation (90-day monitoring period)

Phase 3: Closure Decision

3.4 CAPA Timeliness Standards

4. SPC Interpretation Decision Logic

4.1 Control Chart Selection Flowchart

START: What type of data are you charting?
  │
  ├─ CONTINUOUS (variable) data — measurements in units (mm, kg, °C, psi)
  │   ├─ Are you taking subgroups (multiple measurements per sampling event)?
  │   │   ├─ YES → What is the subgroup size (n)?
  │   │   │         ├─ n = 2 to 9  → X-bar / R chart
  │   │   │         ├─ n = 10 to 25 → X-bar / S chart
  │   │   │         └─ n > 25 → X-bar / S chart (consider reducing subgroup size)
  │   │   └─ NO (n=1, individual readings) → Individuals / Moving Range (I-MR) chart
  │   │         Use when: batch process, destructive testing, slow process,
  │   │         or when each unit is unique
  │   └─ (Verify data normality assumption for variable charts — I-MR is sensitive
  │       to non-normality; consider transformation or use nonparametric alternatives)
  │
  └─ ATTRIBUTE (discrete) data — counts or proportions
      ├─ Are you counting DEFECTIVE ITEMS (units that pass or fail)?
      │   ├─ YES → Is the sample size constant?
      │   │         ├─ YES → np-chart (count of defectives, fixed sample)
      │   │         └─ NO  → p-chart (proportion defective, variable sample)
      │   └─ NO  → You're counting DEFECTS (multiple defects possible per unit)
      │             ├─ Is the inspection area/opportunity constant?
      │             │   ├─ YES → c-chart (count of defects per unit, fixed area)
      │             │   └─ NO  → u-chart (defects per unit, variable area)
      │             └─ (Verify Poisson assumption for c/u charts)
      └─ (Attribute charts require larger sample sizes than variable charts for
          equivalent sensitivity — minimum ~50 for p/np, ~25 for c/u)

4.2 Out-of-Control Response Protocol

When a control chart signals an out-of-control condition, follow this response based on the specific signal:

Rule 1: Point beyond 3σ control limit

Rule 2: Nine consecutive points on one side of the center line (run)

Rule 3: Six consecutive points steadily increasing or decreasing (trend)

Rule 4: Fourteen consecutive points alternating up and down (stratification/mixing)

4.3 Capability Index Interpretation

Cp vs. Cpk Interpretation:

Pp/Ppk vs. Cp/Cpk:

5. Inspection Level Determination

5.1 Incoming Inspection Level Decision Matrix

Inspection Level Assignment:

5.2 ANSI/ASQ Z1.4 Quick Reference

Sample Size Code Letters (Normal Inspection, General Level II):

Switching Rules:

5.3 Skip-Lot Qualification Requirements

Qualification Criteria (all must be met):

1. Supplier is on the Approved Supplier List with "preferred" or "qualified" status
2. Minimum 10 consecutive lots accepted at normal inspection level
3. Supplier's process capability (Cpk) for critical characteristics ≥ 1.33, verified by supplier data AND incoming inspection data
4. No open SCARs against the supplier for this part number
5. Supplier has a certified quality management system (ISO 9001 minimum; industry-specific certification preferred)
6. Written agreement documenting skip-lot terms, reversion criteria, and data submission requirements

Skip-Lot Frequencies:

6. Supplier Quality Escalation Ladder

6.1 Detailed Escalation Process

Level 0: Normal Operations

• Supplier meets scorecard expectations (PPM < threshold, OTD > threshold, SCAR closure on time)
• Standard incoming inspection level
• Quarterly scorecard review
• Annual audit (if risk-based schedule warrants)

Level 1: SCAR Issued

• Trigger: Single significant non-conformance (> $5,000 impact or safety/regulatory concern) OR 3+ minor non-conformances on the same part in 90 days
• Actions:
  • Formal SCAR issued with 8D or equivalent RCA requirement
  • Supplier has 10 business days for initial response (containment + preliminary root cause)
  • Supplier has 30 calendar days for full corrective action plan with implementation timeline
  • Quality engineering review of SCAR response for adequacy
  • Increase incoming inspection level for the affected part number
• Exit criteria: SCAR accepted and closed with verified effectiveness (90-day monitoring)

Level 2: Supplier on Watch / Probation

• Trigger: SCAR not responded to within timeline OR corrective action not effective (recurrence during monitoring) OR scorecard falls below minimum threshold for 2 consecutive quarters
• Actions:
  • Supplier notified of probation status in writing (Quality Manager or Director level)
  • Procurement notified; new business hold (no new part numbers awarded)
  • Increase inspection level for ALL part numbers from this supplier (not just affected part)
  • Monthly performance review calls with supplier quality management
  • Supplier must submit a comprehensive improvement plan within 15 business days
  • Consider on-site quality audit focused on the specific failure mode
• Exit criteria: Improvement plan accepted + 2 consecutive quarters meeting scorecard minimum + no new SCARs

Level 3: Controlled Shipping

• Trigger: Continued failures during watch period OR critical quality escape that reaches customer
• Actions:
  • Controlled Shipping Level 1 (CS-1): Supplier adds additional sort/inspection step with data submitted per shipment
  • If CS-1 ineffective within 60 days: Controlled Shipping Level 2 (CS-2): third-party resident inspector at supplier's facility, at supplier's expense
  • All sort/inspection costs debited to supplier
  • Weekly performance review calls with supplier VP/GM level
  • Begin qualification of alternate source (if not already underway)
• Exit criteria: 90 consecutive days of zero non-conformances under controlled shipping + root cause fully addressed + systemic improvements validated

Level 4: New Source Qualification / Phase-Out

• Trigger: No sustained improvement under controlled shipping OR supplier unwilling/unable to invest in required improvements
• Actions:
  • Formal notification to supplier of intent to transfer business
  • Accelerated alternate supplier qualification (expedite PPAP/FAI/first articles)
  • Reduce business allocation as alternate source ramps up
  • Maintain controlled shipping on remaining volume
  • Ensure last-time-buy quantities cover the transition period
  • Document all quality costs incurred for potential recovery
• Timeline: Depends on part complexity and alternate source readiness; typically 3-12 months

Level 5: ASL Removal

• Trigger: Qualification of alternate source complete OR supplier's quality system failure is fundamental (e.g., data falsification, loss of certification)
• Actions:
  • Formal removal from Approved Supplier List
  • Final shipment received and inspected under 100% inspection
  • All supplier-owned tooling at our facility: disposition per contract terms
  • Our tooling at supplier's facility: retrieve per contract terms
  • Close all open SCARs as "supplier removed"
  • Retain supplier quality file for minimum 7 years (regulatory record retention)
  • Update OASIS (aerospace) or relevant industry databases
• Re-entry: If supplier applies for re-qualification, treat as a new supplier with full qualification process; require evidence that systemic issues were addressed

6.2 Escalation Decision Quick Reference

7. Cost of Quality Calculation Models

7.1 COQ Category Definitions and Tracking

Prevention Costs (invest to prevent defects):

Appraisal Costs (cost of verifying conformance):

Internal Failure Costs (defects caught before shipment):

…(truncated)