OOS Investigation: What a Buyer Should Receive, and When

An email arrives with a single line: the assay came in at 97.8 percent against a 98.0 percent limit, the laboratory has repeated the test, and the repeat passed. The batch is offered for shipment. Whether that email is acceptable depends on something it does not say, which is whether the repeat was a test of the material or a test of the analyst's luck.

An out-of-specification result is a process, not a number. Buyers who treat it as a number end up accepting conclusions they cannot defend, and the failure surfaces later, at the customer's own release step or at an inspection several years downstream.

Two phases, and why the distinction is everything

A structured investigation separates a laboratory phase from a full investigation. The laboratory phase asks one question: is there an identifiable fault in the measurement itself? An instrument calibration failure, a mis-prepared standard, a wrong injection volume, a documented transcription error. If such a fault is found and demonstrated, the result may be invalidated and the test repeated.

The full investigation begins when no laboratory fault is established. It looks at the material: the production record, the process parameters, the raw materials, the equipment history. Its conclusion is about the batch, not about the measurement.

The failure mode buyers encounter most often is a laboratory-phase conclusion reached without a laboratory-phase finding. The repeat passed, therefore the first result was an error. That reasoning is circular. A passing repeat after a failing result is expected roughly half the time in a method with ordinary variability, and it proves nothing on its own.

What the report should contain

A report that can be assessed has a predictable shape. If a supplier's investigation document is missing two or more of these elements, the conclusion is not yet supported.

Element	What it establishes
Original result and the specification it breached	That there was an OOS at all, and against which limit
Hypothesis tested	That a specific cause was proposed, not a general suspicion
Evidence for the hypothesis	Instrument logs, standard preparation records, integration review
Phase reached and conclusion	Whether the batch is still in question
Corrective action	That the cause will not recur on the next batch
Impact on other batches	Whether the same cause affected material already shipped

The last two lines are the ones most often absent, and they are the ones that determine whether the answer is about this batch or about the supply chain.

The invalidation test

An invalidation is legitimate when a defined, documented finding explains the original result and the finding is independent of the result itself. It is not legitimate when the finding is constructed after the fact to reach a preferred conclusion.

The practical test is simple. Ask what would have happened if the repeat had also failed. If the answer is that the same laboratory fault would have been identified, the invalidation is a finding. If the answer is that a different explanation would have been found, it is a justification.

Timing as a signal

An investigation that concludes within a day, with a named fault and instrument evidence, is plausible. An investigation that concludes within a day with no named fault is not. Conversely, an investigation that has been running for six weeks with no phase stated is a different problem: the batch is in limbo and the buyer has no release date.

What a buyer should ask for is the phase, the finding and the evidence, in that order. A date is the least informative of the four.

What to do with the batch

Where the conclusion is a production failure, the batch should not be released, and the useful conversation is about replacement and about whether the same cause affected earlier shipments. Where the conclusion is a properly evidenced laboratory fault, release is defensible, provided the corrective action addresses the measurement rather than the paperwork.

Where the conclusion is ambiguous, the option that costs least is usually a retest from a fresh sample, by a defined method, with the buyer's own laboratory or a contracted one. That is not a substitute for the supplier's investigation. It is a way of deciding whether to keep investing in it.

Why this belongs in a supply agreement

Notification of an OOS result before shipment, and provision of the investigation on request, are contractual matters. Suppliers who accept them in principle rarely have a problem operating them. Suppliers who resist them usually have a process that has not been tested.

For a first shipment from a new source, the clause is worth more than a discount. It is the difference between a quality relationship and a sequence of surprises.

The question behind the question

Most buyers who ask about OOS handling are really asking whether the supplier's quality system notices its own failures. A supplier that reports an OOS before shipment, with an investigation attached, is telling you that the system works. A supplier whose certificates are uniformly perfect is telling you nothing at all, because a perfect record over hundreds of batches is statistically improbable and usually means the failures were resolved before they reached the certificate.

That is not an argument for suspicion. It is an argument for asking, once, early, and watching how the answer is produced.

How to read the investigation you receive

Read the hypothesis before the conclusion. A report that opens with the conclusion and then justifies it is a different document from one that states a hypothesis, tests it and reports what happened. The order of the sections tells you which one you have.

Then read the evidence list. Instrument logs, standard preparation records, integration review, sequence file. If the evidence is described rather than attached, ask for it. A conclusion supported by a description is a claim.

Three related articles on this site make the review faster. The twelve-point COA consistency check catches the arithmetic problems that often accompany an investigation. Reading a result that sits at the limit covers the case where the OOS is marginal rather than clear. And batch-to-batch consistency tells you whether the failing batch was an outlier or the visible end of a drift.

Where the expectations are written down

The handling of out-of-specification results is addressed in inspection guidance rather than in a single harmonised text. The Pharmaceutical Inspection Co-operation Scheme publishes guides that many authorities apply. The European Medicines Agency publishes quality guidance covering deviation management and batch release. The US Food and Drug Administration has published extensively on the same subject, and its guidance documents are public.

You do not need to cite any of them in a supplier conversation. You need to know that the expectation of a documented, hypothesis-driven investigation is not a customer preference. It is the standard.

The three questions to ask on the call

First, which phase has been completed. Second, what was the finding that supported the conclusion. Third, what would you have concluded if the repeat had also failed. The third question is the one that separates an investigation from a rationalisation, and it usually produces a pause.

Ask them in that order and take notes. A supplier who answers all three clearly is telling you the system works. A supplier who answers the first and deflects the other two is telling you something equally useful, earlier and more cheaply than a failed shipment would have.

Why the phase matters more than the conclusion

Two investigations can reach the same conclusion and carry completely different weight. One established a laboratory fault before considering the material. The other considered the material first, found nothing conclusive, and returned to the laboratory. The conclusions read identically. The second one was not an investigation of the measurement, it was a search for an explanation.

The distinction affects what you can do next. If the laboratory phase closed on a documented fault, the batch stands on its own and no further action is needed. If the full investigation ran and found no assignable cause, the batch is not necessarily bad but the cause is unknown, which means the next batch may repeat it.

That second case is the one buyers mishandle. The result is explained as unexplained, the batch is released because it meets specification on retest, and the underlying condition persists. Six months later the same failure appears, and by then the earlier investigation is a paragraph in a file nobody reopened.

Ask which phase, and then ask what the phase concluded about the process rather than about the number. Those are two different questions and only the second tells you what to expect from the next delivery.

The record you should keep

Keep the investigation, not the email that summarised it. The email says the batch is fine. The investigation says what was tested and what was found. Two years later, if the same failure recurs or an inspector asks how a deviation was closed, the investigation is the document that answers, and the email is not.

That is a filing decision made in thirty seconds and worth considerably more than thirty seconds when it matters.

Frequently asked questions

Must a supplier tell me about an OOS result on my batch?

Yes, if the batch has not been released and the result affects a specification attribute. You should receive notification before shipment, not with the certificate. A certificate that reports only passing results, with no mention of an earlier failing result, is a document control failure as much as a quality one.

What is the difference between a laboratory error and a production failure?

A laboratory error is a fault in the measurement, established by evidence and reproduced. A production failure is a fault in the material. The distinction matters because only the second affects the batch. A conclusion of laboratory error that is not supported by a documented, reproducible finding should not be accepted at face value.

How long should an investigation take?

Long enough to test the hypothesis, short enough not to become a way of waiting out the customer. A laboratory-phase investigation is usually days. A full investigation involving production records takes longer. What matters is that you receive the phase, the finding and the evidence, not a date.

Can a batch be released with an OOS result that was invalidated?

Yes, where the invalidation is properly established and documented. The invalidation has to rest on a defined finding, not on the inconvenience of the result. The record of that finding is what a buyer should retain, because it is the part an inspection will ask for.

Should I ask for the raw data?

Ask for the chromatograms, the integration parameters and the sequence file for the affected test. These show whether the result was reviewed or merely reported. A summary table alone does not let you judge whether the investigation was real.

What if the supplier refuses to share the investigation?

Then you are being asked to release a batch on trust. That is a commercial decision, not a quality one, and it should be recorded as such. For a first shipment or a new source, refusal is usually sufficient reason to hold.