Auditing a batch against its COA

By Mohamad Sinno, Charcoal Expert · published 7 September 2026. Not reviewed by the Head of QC — the reviewer byline on our older quality pages carries the date those pages were read, and this one has not been read yet.

Most incoming-inspection guidance is written for parts and chemicals, and the one charcoal version in circulation is a supplier's own playbook quoting its own sample counts. This is the protocol as a buyer runs it, built from the records a charcoal factory actually keeps — ours, named — so you know what to ask for and what a refusal means.

The short answer

Start at the carton, not at the certificate. The batch code printed on the box is the only thing in the transaction that came from the production line and is sitting in your own warehouse; every document should be pulled to match it. A certificate you were sent, for a batch you cannot find on a carton, proves that a test happened somewhere.

The six steps

1. Read the code off the carton

Ours is a printed code in three parts — a four-digit batch production code, two letters identifying the packing team, and one check digit. The four digits are the batch. The two letters are the reason a packing question has a person attached to it. Inner-box marking: not recorded. So if your process depends on marks inside the master carton, agree that separately rather than assuming it.

Take the code from a carton you can touch. Photograph it. Everything below is a request framed by that number.

2. Ask for the batch file, not for "the COA"

Behind that code sits a file for the batch. Ours holds the laboratory certificate of analysis plus the factory's own results for the same batch: ash content, moisture, burning time, drop test, cracks and temperature. Asking for the file rather than the certificate is what separates an audit from a document request — the certificate is one sheet in it.

Two records inside our own process are not published, and we would rather say so than show you a sample page: The drying log and the container loading log are not published, and no sample page of either has been supplied. They exist as working documents; nothing on this site reproduces one. If a supplier hands you an illustrative log, ask which container it was.

3. Ask how the batch was sampled

A figure is a claim about a sample before it is a claim about a shipment. Our sampling is at at random, at least five pieces per 50 kg, at 6 stages — carbonization, screening, grinding, mixing, extrusion, drying — with weight, ash content, moisture, ash colour recorded each time.

Ash colour is the one on that list that does not look like a specification, and it is worth asking about: it is an observation the sampler writes down, not a measured value, and it is the earliest visible sign that something upstream moved.

4. Repeat the two checks you can run yourself

You do not need a laboratory to falsify a batch file. Two of the checks in ours are reproducible in a warehouse:

• Moisture, by drying oven. Ours: drying oven at 120 °c for 5–10 minutes, weighing before and after. That is a quick check, not the constant-mass method — the difference, and when each one is the right tool, is on the moisture page.
• The drop test. Ours, for cubes: The piece is heated until glowing red, held five minutes after heating, then dropped five times from 1.2 m onto a hard tiled surface. It must not break. Hexagonal: three drops from 50 cm. Both are specific on purpose — a drop test without a height, a count and a surface is not a test you can repeat or dispute.

Compare your results against the batch file, not against the grade specification. The specification is a ceiling; the file is what this batch measured, and a batch inside the ceiling but outside its own file is the more interesting finding.

5. Read the method column, and look for the basis

Every value on a certificate was produced by a named method, and a certificate that does not name them is not auditable. On ours, moisture and volatile matter are run to ASTM D1762-84 (Reapproved 2021); ash and fixed carbon are run to the laboratory's own in-house procedure, IK.BPL.7.4, at its points 6.3 and 6.4. That split is normal and it is not a defect — but it means two of the four values on our certificates rest on a method you cannot look up, which is a fact worth knowing before you compare our ash figure against somebody else's.

Then look for the word that is usually missing: the basis. As-received and dry-basis figures are not comparable, and a fixed-carbon number is flattered by several points when it is quoted dry. None of our three certificates states a basis, so we do not publish one — and a supplier quoting a dry-basis figure off a certificate that never said "dry" is quoting something the laboratory did not write.

6. Raise it inside the claims window

Ours is seven days from receipt for quality and quantity claims, after which the goods are deemed accepted; the remedy is replacement or credit on the next order, with compensation by mutual agreement and settled in subsequent shipments. Whatever your supplier's window is, find it before the container lands — an audit protocol whose findings arrive after the deadline is a filing system.

Where AQL fits, and where it does not

Acceptance sampling is the formal version of "how many boxes do I open". It is a real discipline with published standards behind it — ISO 2859-1 and ANSI/ASQ Z1.4 are the two you will meet — and the sample size scales with the lot rather than staying fixed. ANSI, publishing the standard's purpose, describes it as designed to pressure the producer, through the commercial consequence of a rejected lot, into holding a process average at least as good as the agreed quality limit, while capping the buyer's risk of accepting an occasional bad one (blog.ansi.org, 18 February 2026 — ISO's own page is paywalled and we have not read it, so the edition designation is ANSI's).

Two honest limits belong next to that. The first is stated by one of the inspection platforms that sells the method: an AQL plan is a statistical filter and not a guarantee, and it will occasionally accept a lot that is worse than the limit and occasionally reject one that is fine (tryharmony.ai, 14 June 2026). The second is that those standards are attribute-sampling schemes — they count defective units. Ash content is not a defective unit; it is a measurement, and it belongs to the certificate-and-batch-file route above rather than to a sampling table.

We publish no AQL level, no sample count and no process-capability figure of our own, because we do not run an acceptance-sampling scheme and would be inventing one to look rigorous. You will find charcoal suppliers quoting precise sample counts and capability indices; ask which lot those were computed on, and how many lots the average covers.

What a good supplier answer looks like

On the last row, ours: when an in-process check fails — moisture, drop test, cracks, odour — the batch or the failing part of it is rejected and does not ship. We claim no re-drying-and-retest route, because we do not run one.