# Quality decision frameworks

## Containment and disposition

Identify the violated requirement and approved revision before calling a condition nonconforming. Record product and process scope, traceability, measurement uncertainty and confidence. Control suspect stock under the site's authorized procedure and evaluate shipped product promptly. Do not wait for three recurrences, a fixed dollar value or a weekly MRB meeting when safety or mandatory reporting is involved.

Disposition is an authorized decision, not a flowchart's automatic first match:

| Option | Evidence and controls |
|---|---|
| Rework | Approved instructions bring product into full conformance; assess adverse effects and verify required characteristics. |
| Repair | Approved restoration of intended function may leave a deviation; design/customer/regulatory approvals depend on actual requirements. |
| Use-as-is/concession | Document the specific permitted deviation, scope, time/quantity limits, qualified risk assessment and all required approvals. Functional clearance alone does not establish full acceptability. |
| Return | Preserve custody and evidence, especially for suspect counterfeit material; authorize RMA, transport and contractual costs. |
| Scrap | Authorized, traceable disposal/destruction with safety, environmental and evidence-retention requirements. |

No universal 60%-of-replacement cutoff determines rework, and business urgency cannot authorize a concession. Use actual incremental costs and schedule effects only among technically/regulatorily permissible options. Supplier charges, ASL changes and external reporting are separately authorized.

## Current requirements register

For each obligation capture jurisdiction, product/entity scope, current source/standard edition, exact requirement, awareness/trigger date, calculated deadline, owner and evidence of completion. Distinguish regulatory periods, customer commitments and internal targets. Never delay a report until root cause is proven.

FDA revised device Part 820 to [QMSR effective 2 February 2026](https://www.fda.gov/medical-devices/postmarket-requirements-devices/quality-management-system-regulation-qmsr). Map current requirements through the incorporated ISO 13485:2016 and FDA-specific provisions; do not use retired QSR paragraph numbers as present law. FDA device MDR under Part 803 and corrections/removals under Part 806 are separate assessments; manufacturer, importer and user-facility duties differ. Qualified regulatory personnel must evaluate the actual event and awareness trigger; not every complaint or calibration issue is automatically reportable.

IATF, aerospace, ISO and customer-specific duties require the actual applicable text. Do not claim that every process/tool change requires full PPAP/FAI, every suspect certificate requires a specific external report, or every concession has the same approval chain. Drug/supplement cleaning and cross-contamination issues are not device-QMSR scenarios.

A [Form 483](https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/inspection-references/fda-form-483-frequently-asked-questions) records inspectional observations, not a final agency finding. Acknowledge receipt, investigate accurately and distinguish any evidence-supported disagreement from admissions. Verify the current response instructions and promptly target the applicable response period; “15 business days” must not be mislabeled a universal statutory requirement to complete all corrections. Unfinished work needs accurate status, responsible owners and justified dates.

## Root cause and CAPA

Use a precise problem statement: what, where, when, how much, requirement and actual result, affected versus unaffected comparisons. Validate the measurement system; examine personnel, equipment, material, methods and environment. Avoid blaming an individual without evidence or forcing every investigation into exactly five linear questions.

Fishbone and Five Whys produce hypotheses, not proof. FTA may be qualitative; quantitative gate calculations require suitable probabilities and treatment of dependencies/common causes. 8D structures teamwork and containment but does not independently prove cause. Prefer feasible durable controls that address verified causes; no control class has a universal 30%, 90% or 100% effectiveness rating.

Set CAPA initiation criteria according to risk, trend and actual QMS requirements. A single severe event may warrant action; several trivial events do not automatically prove a systemic cause. Define implementation acceptance, risk of adverse effects, planned effectiveness sample/exposure and statistical decision before collecting results. Zero failures in a small or irrelevant sample is not evidence of zero risk. Consider production mix, seasons and failure cycles, not a fixed 90-day or three-lot minimum.

If recurrence occurs, assess cause, scope and continuing containment. Reopen or create a linked record according to the controlled QMS, preserving prior decisions, effectiveness evidence and immutable audit history. Supplier removal or a changed process does not erase outstanding product risk or justify administrative “effective” closure. Retention follows actual product, legal, contract and QMS requirements, not a universal seven years.

## SPC and capability

Select the chart for data generation and rational subgroups: X-bar/R or S for suitable grouped measurements, I-MR for individual readings, p/np for nonconforming units and c/u for nonconformity counts with appropriate exposure. Check dependence, distribution, opportunity, subgroup size and measurement limitations; no single minimum sample size applies to every attribute chart.

Control limits describe process behavior; specification limits are acceptance requirements. Never set control limits equal to a desired reliability threshold or adjust limits simply to hide signals. A stable process can make nonconforming or functionally inadequate product. A control-chart signal is evidence to investigate under the approved reaction plan, not proof of a specific cause.

Do not mix named rule sets. [NIST documents the conventional Western Electric rules](https://www.itl.nist.gov/div898/handbook/pmc/section3/pmc32.htm), which include a point beyond 3 sigma, two of three beyond 2 sigma on the same side, four of five beyond 1 sigma on the same side, and eight consecutive on one side. Nine-on-one-side, six trending and fourteen alternating belong to a different commonly used set (Nelson). Document the chosen set, treatment of ties and false-alarm effects; software defaults vary. Investigate unsafe/out-of-spec conditions even without a chart signal.

For two-sided limits, Cp = (USL − LSL)/(6σwithin) and Cpk = min((USL − mean)/(3σwithin), (mean − LSL)/(3σwithin)); Pp/Ppk use the selected overall variation estimate. State estimator, data window, sample size and uncertainty. One-sided, bounded, non-normal or unstable processes need appropriate methods. A high Cp with low Cpk indicates centering matters but does not prove variation cannot also improve. Fixed Cpk values such as 1.33/1.67 are examples of possible customer targets, not universal IATF pass rules, release authority or a defect-free guarantee. Do not infer future safety or “Six Sigma” solely from one estimated index.

## Acceptance sampling

Use the approved licensed standard/table and customer plan. [ASQ explains Z1.4](https://asq.org/quality-resources/z14-z19) as a system of attribute plans with switching rules. General level I/II/III, normal/tightened/reduced severity, single/double/multiple plan and AQL are distinct inputs. Do not equate Level III with tightened inspection or treat 100% inspection as perfect detection.

Record lot definition and size, randomization, characteristic/defect class, units versus defect counts, standard edition, inspection level, scheme/plan type, AQL or consumer-risk criterion, code letter, table/arrow direction, sample size, Ac/Re and documented switching history. Have a qualified reviewer verify the lookup. This catalog intentionally provides no reconstructed standard tables: the previous illustrative Ac/Re values were unreliable.

A zero-acceptance plan is not “AQL = 0” and does not ensure zero defects in accepted lots. Acceptance sampling estimates neither exact lot defect rate nor safety of every untested unit. Approved critical-characteristic controls may require different methods. Reduced/skip-lot eligibility and discontinuation depend on complete standard conditions and authorized decisions; a streak of ten accepted lots alone is insufficient.

## Supplier and cost management

Base supplier response periods, controlled shipping, auditing and alternate-source qualification on actual agreements, product risk and verified effectiveness. Qualification or a certificate does not establish every lot conforms. Do not remove a sole source, issue debits, change business allocation or make a supplier's costs binding from a generic escalation ladder.

Track prevention, appraisal, internal failure and external failure costs without double counting. Distinguish actual costs, uncertain exposure and projected savings; validate investment assumptions and positive denominators. No universal $10–$100 return per prevention dollar or quality-cost/revenue target is asserted.

## Required handoff

Provide the evidence index, scope and uncertainty, containment status, proposed technical options and approval needs, current requirements/deadlines, causal hypotheses and tests, CAPA/effectiveness plan, costs with assumptions, owners and next checkpoints. No checklist alone is compliance certification.