The sixth marginal lot

A raw material result exceeds its requirement before a scheduled campaign. The immediate event can be contained and evaluated. The harder question is whether this is an isolated failure, the sixth similar event from one source, a drifting method, an unsuitable requirement, or a pattern of concessions that has become normal.

An individually defensible disposition can still be a weak system decision when prior events, supplier performance, affected uses, downstream controls, regulatory commitments, and cumulative risk are missing from the review.

Clear failures are easy. Media arrives contaminated. Reject it. A bioreactor crashes mid-run. Scrap the batch. Nobody argues. The hard decisions are the marginal ones. Endotoxin at 0.35 EU/mL when the spec says 0.25. Titer at 1.8 g/L when acceptance is 2.0. Aggregation at the edge of the distribution.

These decisions reveal how the quality system behaves. One evidence-backed concession may be reasonable. Repeated concessions against the same requirement can normalize failure. But many quality systems cannot show which is happening because each NCMR lives in its own record.

The FDA sees it differently. They don't audit individual NCMRs. They audit the aggregate. They ask for your concession rates by supplier, by material, over time. They want to see whether you're improving or drifting. If you can't produce that picture in minutes, you've already given them the answer.

Seal treats every disposition decision. Use as is, rework, scrap, return. As a data point in a trajectory. When the MRB convenes for the sixth endotoxin exceedance, they see it's the sixth. Not because someone remembered, but because the system connects every decision about the same failure mode, the same supplier, the same specification. The question changes from "is this batch acceptable?" to "what is the pattern of our decisions telling us?"

Your cross-functional review process works. Engineers assess technical risk. QA checks compliance. Manufacturing flags handling concerns. Each decision is defensible in isolation.

The problem is that each MRB sees one event. The board that reviewed the sixth endotoxin exceedance didn't know it was the sixth. Because their QMS showed them a single NCMR, not a trajectory. They made a reasonable decision with incomplete information.

Seal puts the pattern in front of the MRB before they deliberate. When reviewers open NCMR-2024-047, they see every prior NCMR for that supplier, that material, that specification. The trend line is right there. The concession rate over the last twelve months is right there. The discussion changes. Not because the people are different, but because the information is.

This is what a connected platform makes possible. Your current QMS captures MRB decisions. Seal connects them to supplier history from incoming inspection, to batch outcomes from MES, to stability data from LIMS. The MRB doesn't just know about this event. They know what happened last time they accepted the same deviation.

A serious downstream consequence rarely begins with one obviously reckless decision. It can emerge when several reviewers make similar decisions independently over time and nobody can see the cumulative pattern. If a later signal changes the risk assessment, the affected batches and shipments must be found immediately.

That's because in a standalone QMS, the NCMR is a document. It references lot numbers, but those lot numbers don't link to batch records in your MES, which don't link to distribution records in your ERP. The forward trace is a manual investigation across disconnected systems.

In Seal, the NCMR, the batch record, and the distribution record are on the same platform. The moment an NCMR is opened, automatic inventory holds block the material from being issued to any batch. Not a tag that can fall off, but a system-level lock that a barcode scan can't override. When the NCMR is dispositioned, the links are already there. The trace from "NCMR-2024-001, raw material lot RM-2024-0847" to "all affected drug substance batches and clinical shipments" is a single query.

When the FDA asks "which products were affected?"—you show them in minutes, not days. That speed isn't a feature of the QMS module. It's a consequence of the QMS, MES, and inventory living on one platform.

Most management reviews look at NCMR counts. Open vs. closed. Aging. Maybe a Pareto chart of root causes. This tells you almost nothing about whether your quality is improving.

The questions that matter: Are concession rates increasing? Which suppliers are generating the most nonconformances, and is it getting better or worse? Are the same specifications failing repeatedly. And if so, is the specification wrong or is the process drifting? When a CAPA closes, do the nonconformance rates for that failure mode actually decrease?

Seal surfaces these patterns in real time. When a supplier's rejection rate crosses a threshold you define, the system flags it before the next MRB. When the same specification fails for the third time in six months, it's visible to everyone reviewing the NCMR. Not buried in a filing cabinet. When a CAPA claims to have fixed a problem, the nonconformance data shows whether it actually did.

The sixth similar exceedance should not be presented as the first. Connected data turns recurrence into an expected escalation, with thresholds and actions defined before the next review.

Item, material or product lot, component, batch, equipment, process step, sample, location, supplier, requirement and version, observed condition, quantity, unit, time, detector, evidence, and immediate risk form the initial record.

Suspected, confirmed, invalidated, duplicate, and related-event states remain explicit. Correcting an entered value never erases what was first observed.

Inventory holds, equipment or area restriction, process stop, electronic block, physical segregation, label, access control, communication, sample preservation, and temporary instruction can be initiated before the full evaluation.

Each control retains scope, owner, start, verification, exceptions, release authority, and end. A status field alone is not proof that affected material could not be used.

Forward and backward trace considers source lots, sibling containers, supplier shipments, samples, equipment, time windows, batches, intermediates, finished lots, packaging, serials, customers, sites, studies, and retained inventory.

Population versions preserve why items were added or excluded. A later laboratory, supplier, process, complaint, or stability signal can reopen and expand the assessment.

Source, type, severity, product and patient impact, safety, data integrity, regulatory impact, detectability, recurrence, scope, released or distributed state, and investigation need determine classification.

Rules identify required reviewers, response times, escalation, external notifications, investigation, CAPA consideration, and disposition authority without replacing quality judgment.

Requirement validity, measurement reliability, sampling, material identity, history, supplier data, process capability, downstream steps, specification rationale, product impact, intended use, stability, validation, regulatory commitments, and alternative dispositions are evaluated as evidence.

Uncertainty and assumptions remain visible. The record distinguishes what is known, inferred, missing, and still required before disposition.

Quality, manufacturing, engineering, laboratory, supply, regulatory, safety, medical, design, supplier quality, and other roles participate according to the event and product.

Agenda, evidence reviewed, quorum, conflicts, questions, votes or concurrence, dissent, decision, rationale, conditions, signatures, and escalation remain traceable.

Accept under concession, sort, screen, rework, reprocess, repair, relabel, downgrade, return to supplier, scrap, destroy, or other approved paths define scope, instructions, prerequisites, risks, approvals, quantities, sites, owners, and completion evidence.

No inventory state changes merely because the MRB selected an option. Release or final closure follows verified execution and reconciliation.

The specific failed requirement, quantity and population, intended use, scientific and technical rationale, risk, downstream controls, prior concessions, cumulative exposure, regulatory impact, customer or authority approval where applicable, conditions, expiry, and approvers define the concession.

A concession does not revise the requirement or create precedent. Recurrence thresholds can require escalation, supplier action, specification review, CAPA, or block further concession.

Approved instructions, authorized people, equipment, materials, parameters, records, sampling, acceptance criteria, validation or engineering basis, traceability, reconciliation, deviations, and post-execution inspection or testing govern rework and repair.

The original nonconformance stays visible through execution. Successful reinspection proves the disposition completed; it does not erase the event.

Container and quantity scans, moves, destruction witnesses, certificates, supplier return authorization, shipment, rework records, test results, labels, inventory adjustments, exceptions, and independent verification complete the plan.

Expected, executed, remaining, lost, sampled, transformed, returned, and destroyed quantities reconcile before closure.

Incoming and latent supplier-attributed events connect to legal entity, manufacturer, site, material or service scope, purchase order, receipt, supplier lot, quality agreement, prior performance, notification, response, SCAR, change, and requalification.

Supplier attribution remains evidence-based and can change during investigation without breaking the original trace.

Material, supplier, site, process step, equipment, requirement, failure mode, defect code, cause, disposition, product, and time can identify related events. Confirmed relationships preserve reviewer rationale; likely matches remain suggestions until assessed.

Threshold rules can initiate escalation before the next MRB convenes.

For corrective or preventive controls, Seal defines the monitored population, period, expected event rate, acceptance threshold, confounders, data source, reviewer, outcome, and next action.

Closure based only on task completion is distinguishable from evidence that recurrence or concession rate actually declined.

Event rate, supplier defect rate, concession rate, repeated failure mode, aging, containment time, disposition mix, rework success, scrap cost, recurrence, and effectiveness use defined populations and exclusions.

Teams can drill from every trend to the events, decisions, affected lots, and physical outcomes behind it.

A nonconformance begins with an item, material, component, product, service, or output that does not meet a requirement and needs control and disposition. A deviation begins with a departure from an approved instruction, process, parameter, or expected execution and needs investigation.

One event may create or relate to the other. Seal connects them while preserving the different containment, evaluation, and decision responsibilities.

The NCMR is connected to inventory status and place, samples and results, supplier and requirement history, batch use, distribution, rework records, destruction, CAPA, and later performance.

That is what turns an MRB decision from an approved document into a verified control over the real affected population.