Corrective Action Request (CAR)

A Corrective Action Request (CAR) is a documented request for a responsible person or organisation to investigate a quality problem, address its cause and provide evidence for review. CAR is industry terminology, not a defined regulatory term. Its operational meaning comes from the organisation’s quality procedures and, for supplier requests, the agreed customer–supplier process.

The request identifies the problem, assigns accountability and establishes what response is expected. It turns an observation into an explicit assignment: who will investigate, what issue they will address and what evidence the reviewer expects to receive. Issuing a CAR does not establish a root cause or demonstrate that an action worked. Those conclusions depend on subsequent investigation and evaluation, not on the existence of the request.

Typical origins include a QA audit finding, recurring nonconformance, complaint trend, manufacturing deviation or supplier problem such as a receiving discrepancy. Keep the originating evidence identifiable rather than replacing it with a summary. A reviewer needs to trace the request back to the observation, affected material or process, and supporting records. Describe what was observed without presenting an untested explanation as the cause.

A supplier corrective action request is commonly abbreviated SCAR. It directs the request to a supplier through the agreed customer–supplier process. The supplier section addresses responsibilities for that exchange; the starting purpose remains a clear problem statement and an accountable request for an evidence-based response.

In industry practice, a CAR requests work; it is not necessarily the record of the original event, the action performed or the broader CAPA record. The distinctions below describe common operational uses, not a universal record hierarchy.

Sites commonly distinguish a deviation from a nonconformance through their procedures; others use overlapping categories. Correcting an incorrect label resolves the immediate defect. Changing the selection control that allowed that label addresses its cause. Neither the category name nor completion of the correction establishes that corrective action succeeded.

A CAR can link to an existing investigation or CAPA rather than duplicate its analysis. Relationships can be many-to-many: related findings may share an investigation, while a broader CAPA may coordinate supplier requests and internal actions. Preserve each finding’s identity and the evidence supporting its resolution. Escalation depends on the site’s criteria, not the abbreviation on the form; CAPA effectiveness evaluation belongs with the coordinated action plan.

Corrective action request is an industry-practice label, not a defined regulatory term. A site's choice to call a record a CAR does not establish which legal framework applies. Keep the applicable obligation or guidance separate from the local procedure that names, routes and closes the request.

In practice, procedures commonly map each CAR to its initiating record and identify the basis for the requested response: an applicable provision, a guidance recommendation or an internal control. This makes review more precise than attaching a generic regulatory reference to every request. For device-specific context, see FDA QMSR and the distinction between QMSR and ISO 13485.

A useful CAR field model connects the observation, requested response and acceptance decision without turning the request into a duplicate investigation file. The following is a suggested practice model, not a prescribed regulatory form. Use stable identifiers and explicit relationships so reviewers can follow each conclusion back to its evidence.

Describe the applicable expectation and observed condition before proposing a cause; keep hypotheses separately labelled. Retain authoritative investigations, CAPA records and change documents in their controlled locations, linking the relevant revision rather than copying competing versions. Store acceptance_criteria, verification_method and exposure_basis with each effectiveness check; reference those checks at closure. A target date is not an acceptance criterion: it schedules work but does not establish whether the action succeeded.

Companion data model

Suggested joins, with arrows pointing from foreign keys to referenced keys:

CAR.origin_record_id → OriginRecord.record_id SupplierResponse.car_id → CAR.car_id Action.car_id → CAR.car_id Action.supplier_response_id → SupplierResponse.response_id, when applicable EffectivenessCheck.action_id → Action.action_id ClosureDecision.car_id → CAR.car_id ClosureCheck.decision_id → ClosureDecision.decision_id ClosureCheck.check_id → EffectivenessCheck.check_id

Here, ClosureCheck links a decision to every check considered. Action and check identifier lists in the field table represent linked records, not comma-separated text. Use junction records where actions are shared across requests or checks assess multiple actions. Preserve supplier response revisions with their evidence references, and identify the response supporting each commitment. Record the evidence provider separately from the accepting reviewer; approved_by identifies the human authorising closure, not the person merely uploading the response.

This illustrative industry-practice workflow uses evidence gates rather than treating a response as closure. Local procedures determine routing and approval authority; incomplete evidence sends the request back to the responsible owner.

Intake and acknowledgement → Containment and scope → Investigation or investigation linkage → Action-plan review → Implementation → Effectiveness evaluation → Closure decision

Worked case: dispensing configuration

Assume an internal pharmaceutical dispensing discrepancy is associated with a balance-unit configuration; mass observations are recorded in grams. Link the CAR to the deviation, work order, dispensing entries and batch evidence. Preserve the as-found configuration and examine whether other work orders used the same setup before narrowing scope.

Treat configuration failure as a hypothesis until controlled challenges support it. An operator-error explanation remains untested without evidence distinguishing user actions from system behaviour; retraining alone does not demonstrate resolution.

If controlled challenges confirm the configuration hypothesis, link the configuration change and supporting instructions through the site's applicable change-control process. Approval and execution records demonstrate implementation. Separately, test correct operation across the intended settings against predefined criteria and retain the observations as effectiveness evidence. Propose closure only when those results and the assessment of other work orders support it; otherwise revisit the cause, action or scope.

CAR triage commonly considers potential product impact, affected scope, detectability, recurrence and uncertainty. Assess which batches, products, sites or customers may be affected; whether existing controls would detect the problem before use; and what remains unknown. Severity labels are site-defined categories, not universal regulatory classifications. Record the reasoning, not just the label, in the CAR or linked quality risk register.

Separate immediate containment priority from the investigation and corrective-action schedule. An urgent need to isolate potentially affected material does not establish how long a defensible investigation will take. Set distinct milestones for containment confirmation, investigation findings, action delivery and effectiveness review through the site's procedure or supplier agreement.

Assign an accountable CAR owner and select reviewers for the issue: quality, manufacturing, engineering, laboratory, purchasing or supplier quality. Use supplier risk management to inform supplier-facing routing. Route cross-site or cross-product patterns to an owner with authority to evaluate the broader process, rather than letting local requests close independently.

For an extension, record the justification, current containment, residual risk, revised commitment and approving role. A changed due date alone does not explain whether continued operation remains acceptable.

Define reassessment triggers: expanding scope, ineffective containment, new complaints, contradictory evidence or repeated missed commitments. Reassessment commonly revisits priority, ownership, containment and escalation level; it is not merely another reminder to the same assignee.

A supplier corrective action request (SCAR) commonly packages the affected material or service, purchase and delivery references, observed problem, supporting evidence and requested response milestones. Link the originating receiving discrepancy so the supplier can distinguish the documented observation from an initial suspicion about its cause.

Keep separate supplier and customer owners. The supplier provides its investigation and action evidence; the receiving organisation evaluates whether the response addresses its affected scope. Use supplier qualification history and the applicable quality agreement to interpret commitments, response expectations and escalation routes rather than assuming every supplier follows the same process.

Review the evidence behind the response

A practical acceptance review tests completeness, causal reasoning, action-to-cause linkage, implementation evidence and the proposed effectiveness method. Ask whether the finding also affects other supplied materials or production locations. A completed supplier template is an organised response, not sufficient evidence by itself: each conclusion needs support that the customer can evaluate.

Worked case: outdated allergen declaration

Assume a food manufacturer receives packaging carrying an outdated allergen declaration. Track the potentially affected scope by deliveries and material lots. Before selecting a cause, compare the customer's approved artwork, records of revision transmission and acknowledgement, and the supplier's production setup. These comparisons distinguish an incorrect instruction from a transfer failure or use of an obsolete production file.

Replacement packaging addresses the affected delivery. If the investigation confirms a revision-transfer failure, the corrective action instead targets the transfer and approval process. Evaluate the supplier's implementation evidence against the customer's approved artwork, then use subsequent relevant deliveries for an effectiveness check. Select deliveries that exercise the changed process; unchanged repeat orders may not test revision transfer.

If the response is rejected, record specific evidence gaps and the resubmission request, including what would resolve each gap. Preserve response versions, attachments and the customer's review decision so acceptance traces to the actual response reviewed—not merely the supplier's latest file or the date a form arrived.

As industry practice, define the effectiveness question before reviewing results: which failure mechanism should change, which observations can test that expectation, and what constitutes failure or insufficient evidence? Choose exposure that exercises the mechanism—relevant production runs, supplier deliveries or challenged operating conditions. Elapsed time alone says little when the process has remained idle.

No detected recurrence is not proof that the changed control addresses the cause. Detection coverage, operating conditions and opportunities for failure affect that conclusion. Record these limits; do not attach unsupported statistical-confidence claims to a closure decision.

Worked case: device packaging seal width

Assume a medical-device packaging process has variable seal width, measured in millimetres, and investigation links the variation to an adjustable setup feature. The action adds a positive stop and locking adjustment. Installation records and setup instructions support implementation completion; they do not establish effectiveness.

Before testing, define seal-width acceptance criteria, the intended operating range, challenge conditions and relevant subsequent production exposure. Use the approved parameter tolerance band, rather than limits selected after seeing results. Challenge the changed setup across that range, then review measurements from relevant production runs. Passing results under the planned conditions support an effectiveness decision within that scope. A criterion breach triggers failure assessment; insufficient production exposure leaves the check pending, not passed. The reviewer records the evidence and reasoning supporting closure.

Metrics with explicit counting rules

Let P be the reporting period and T its closing date. Use these cohort rules consistently; report exclusions separately. Shares are fractions, not counts.

Stratify by supplier, product or severity where differences affect interpretation; do not pool unlike exposure units.

Treat CAR integration controls as implementation practices, not additional regulatory requirements or V5 capability claims. In MES–QMS integration, map the CAR to authoritative deviation, complaint, audit, supplier, batch, laboratory and change-control records using stable identifiers rather than copied titles.

Maintain an ownership map naming the authoritative system for each status and evidence item. For example, the QMS may own CAR review status, the supplier portal the submitted response, MES the batch execution record, and LIMS the laboratory result. Record the source identifier, revision and transfer status at each handoff; distinguish successful receipt from evidence acceptance.

• Duplicate imports or retries: retain the external identifier and check it before creating a request. Associate a retry with the existing CAR and preserve its original history rather than creating competing requests.
• Failed transfers or missing attachments: expose the exception to a named owner, reconcile against the originating record and keep affected review decisions pending until the relevant evidence is available. A delivered message does not establish attachment completeness.
• Revised supplier responses or investigation conclusions: retain the earlier version, identify dependent reviews and reassess them against the revision. Do not let stale approval evidence represent a materially changed response.

Keep evidence assessment and approval with authorised people. Use an audit trail review workflow to inspect revisions and recovery history before accepting the reconciled record.

Use these practice-based checks to test decision quality, not merely form completion. Apply the terminology from the earlier comparison section and the evidence structure from the field section.

Implementation checklist

• Agree terminology and escalation criteria across participating teams.
• Name owners and approvers, including delegated coverage.
• Control evidence references so reviewers retrieve the intended version.
• Define supplier-response acceptance rules and who applies them.
• Establish extension and reopening routes with recorded decisions.
• Define metric denominators and inclusion rules.
• Test recoverable record handoffs, including failed transfers and reconciliation.

Review checklist

Select a completed request. Trace its originating observation through scope, investigation, action evidence, effectiveness result and approval. Examine an overdue or reopened request to test the exception path rather than assuming the normal workflow covers it.

Check whether management review identifies recurring causes, ineffective controls or resource constraints, rather than reporting only closed-request volume. Look for decisions and assigned follow-up where those patterns appear.

The practical usability test: an independent reviewer can understand the decision and locate its supporting evidence without reconstructing the history from email.