Complaint Handling for APIs: What a Closed Loop Looks Like

Three pallets arrive at a formulator's warehouse. The powder is white at the top of each drum and faintly cream at the surface. The quality team raises a complaint with photographs. Two days later the reply arrives: slight colour variation is within the normal range for this product.

The sentence may be true. It is still not an investigation, and the difference between those two things is the subject of this article.

Seven parts of a closed loop

A complaint is closed when the customer can see that the event was contained, explained and prevented. That requires seven elements, and a reply missing any one of them leaves the record open in the customer's system even if the supplier considers it finished.

#	Element	What it answers
1	Acknowledgment	Received, owned, named contact
2	Containment	What happened to the remaining stock, at both sites
3	Batch traceability	Which batches share the cause, where they went
4	Investigation	What was examined, what was found, with evidence
5	Root cause	Why it happened, not what it looked like
6	Corrective and preventive action	What changes, and where else it applies
7	Effectiveness check	How you will know the action worked

Most replies contain 1, 2 and a version of 4. Elements 3, 5 and 7 are the ones that separate a supplier with a quality system from a supplier with a complaints inbox.

Containment before explanation

Containment is the part customers care about first, because it decides whether they can keep producing. If three pallets are affected, the immediate questions are whether the remaining stock in the supplier's warehouse has the same issue, whether material already in transit is affected, and whether the customer should quarantine what they have.

A supplier who opens with an explanation and closes with "please advise if you need anything further" has inverted the order. The customer needs to know what to do with the material in front of them before they need to know why it looks the way it does.

Containment is also the fastest way to discover the scale of the problem. If the same appearance appears in four batches over three months, the complaint is not about a pallet. It is about a process that has been drifting since a change nobody flagged.

Appearance complaints need a reference, not an argument

Colour is the most common complaint in this product family and the least well specified. Most certificates carry an appearance line that reads "white to off-white powder", which is a range wide enough to contain the disagreement rather than to resolve it.

The fix is procedural and it belongs at qualification, not at complaint. Three options work.

A retained reference standard. A physical sample of the approved appearance, sealed, dated, and held by both parties. The comparison is then against an object rather than a memory.

A colour grade. A defined scale, with a stated viewing condition. Lighting matters more than most people expect: the same powder looks different under daylight, cool white fluorescent and a phone flashlight.

A photograph protocol. Defined light source, defined background, defined distance, and a signed copy in the qualification file. Crude, and much better than nothing.

Without one of these, the supplier says "within normal range" and the customer says "not what we received last time", and both are describing a real observation with no instrument to arbitrate it.

Root cause is not a restatement of the complaint

"The colour variation is due to slight oxidation at the surface of the drum" is a mechanism, and a mechanism is closer to a root cause than "colour variation was observed". It is still not a root cause, because it does not say why oxidation happened in this batch and not the previous four.

A root cause answers a why-chain that ends somewhere actionable: the drum was closed at a higher headspace humidity because the packing area's dehumidifier was out of service, and the out-of-service status was not on the batch record. That ends at a maintenance and documentation failure, which can be fixed.

Two techniques are enough for most of these investigations. The five-whys, applied honestly rather than performed. And a comparison of the affected batches against the unaffected ones on every process parameter recorded, which usually shows the difference in one column.

Impact assessment: the part that protects other customers

The customer who complained knows about their own pallets. They do not know whether the same cause affects material shipped to somebody else, and they are entitled to ask.

An impact assessment covers three dimensions: other batches from the same period, other batches made with the same equipment or the same operator shift, and batches already shipped to other customers. Where the answer is "three batches, two customers, both notified", the complaining customer's confidence goes up rather than down.

Where the answer is "only your batch", and the investigation cannot explain why, the assessment is not finished.

Effectiveness, and how to prove it

A corrective action without an effectiveness check is a statement of intent. The check needs a measure and a period: three consecutive batches with surface colour matching the reference standard, verified at release by the same person who signed the original complaint response.

This is also where the loop closes for the supplier's own benefit. If the check fails, the action was wrong and the investigation reopens with better information than it had the first time.

What to do with a weak reply

Do not escalate. Ask three questions in writing: what was contained, what the root cause was, and what evidence supports it. The answers sort suppliers quickly.

A supplier with an investigation replies within a day with a document. A supplier without one replies with a longer version of the original sentence, usually adding the phrase "well-established process". That phrase is worth noting in the supplier review, because it is a claim about the past offered in place of evidence about the present.

Where this leaves a supply-chain partner

We answer complaints in the seven-part shape, containment first, and we say when an investigation is still open rather than closing it early to look responsive. An open investigation with a date is more useful to a customer than a closed one with no evidence, and it is considerably easier to defend at an audit.

If your supplier's last complaint reply was two paragraphs and no document, the next quality agreement is the place to fix it.

The first ten days

Most of the outcome of a complaint is decided before any laboratory work is finished, in the first ten days, by whether the response gives the customer something to act on.

Days one and two. Acknowledgment with a named owner and a reference number. This is administrative and it is the step most often skipped, because the person who receives the complaint is not the person who can answer it.

Days two to four. Containment. Is the remaining stock at the supplier's site affected? Is material in transit affected? Should the customer quarantine what they hold? These are answerable from records without waiting for an investigation, and they are what the customer needs first.

Days four to seven. A preliminary investigation summary: what was examined, what the batch records show, whether other batches share the characteristic, and what the likely mechanism is. Preliminary means preliminary. It does not mean a conclusion dressed as one.

Days seven to ten. A plan with dates for the root cause and the corrective action.

The ICH Q10 model of corrective and preventive action expects the loop to close with an effectiveness check, and the PIC/S guides show how inspectors test whether that happened. The FDA guidance library is useful for the investigation write-up itself, which in this industry is read by people who were not present and who will judge it on whether the logic holds.

Recording a complaint so it survives an audit

A complaint record has four readers: the customer who raised it, the investigator, the auditor two years later, and the next supplier review. A record written for the first two is usually unreadable to the last two.

Four fields do most of the work. The observation, in the customer's words, quoted rather than paraphrased. The batch and quantity, with the distribution record. The investigation, with the data attached rather than referenced. And the conclusion, with the reasoning that connects the data to the action.

The failure mode is a record that says "investigated, no abnormality found" and closes. It may be true. It is not auditable, because the next reader cannot tell what was examined. A record that lists the parameters checked, the batches compared and the results found, including the ones that showed nothing, is auditable and takes twenty minutes longer to write.

This is the same discipline that makes an OOS investigation usable to a customer, and the same discipline behind the certificate consistency check, which catches many complaints before they are raised. Where the complaint is about appearance rather than a number, the resolution usually turns out to be a clause missing from the quality agreement rather than a fault in the batch.

Frequently asked questions

What is the difference between a correction and a corrective action?

A correction deals with the material: replace the pallets, re-test the batch, credit the invoice. A corrective action deals with the cause, so that the next batch does not repeat it. Most supplier replies contain the first and stop. The second is what the customer's CAPA system is actually tracking, and its absence is why the same complaint recurs twelve months later with a different batch number.

How fast should a supplier respond?

Acknowledgment within two working days and a preliminary response within ten is a common contractual shape. The preliminary response should contain containment and an investigation plan, not a conclusion. A conclusion delivered in forty-eight hours is usually a conclusion reached before any investigation started.

What if there is no specification limit for appearance?

Then the specification has a gap, and the gap is the real finding. Colour and appearance complaints are unresolvable without an agreed reference: a retained standard, a colour grade, or a photograph taken under defined lighting and signed by both parties at qualification. Without one, the argument is about perception and no laboratory result will settle it.

Does a complaint always need a CAPA?

No. A single event with a clear correction and no systemic cause can be closed with a correction and a note. The error is the opposite: closing everything as minor to keep the metrics clean, which is visible in the pattern even when each individual record looks defensible.

Should the supplier tell us about complaints from other customers?

Not by name, but yes in substance, if the issue could affect material already shipped to you. An impact assessment that covers only your batch is incomplete. The question is whether other batches share the cause, and a supplier who has not asked that question has not finished the investigation.

What do we do if the reply is a paragraph of reassurance?

Ask three questions in writing: what was contained, what the root cause was, and what evidence supports it. A supplier with an investigation can answer in a day. A supplier without one will answer with a longer paragraph of reassurance, and that is useful information for the next supplier review.