Diagnose a charcoal fault from the symptom

Start with what you can smell, see or hear, not with what you think caused it. Every diagnostic here ends with a test you can run on stock you already hold, without a laboratory and without your supplier’s cooperation.

§1 · How to use this section

Fault-finding on charcoal usually begins with a theory. A brand owner decides the coal is poor quality and asks the supplier to confirm it, which no supplier will. The symptom is better evidence: the character of a smell, the moment in the burn it appears, and whether it varies across a container each point somewhere specific, and at different people.

So the table is keyed on what you noticed rather than on what is wrong. What decides who pays is usually not which fault it is but when you first noticed it, which is why the middle column keeps saying so.

§2 · Symptom triage

Every row now has a page behind it. Where a symptom has more than one cause, the page it sends you to carries the differential that separates them rather than picking for you — which is the only honest thing to do when the same smell has two causes.

§3 · The tests you can run today

Seven tests, and they do not all stand on the same evidence. Five are our own methods and two were never ours. Four now have a result of ours published against them, which is the split below — because a method we only propose is worth less to you than one we have answered in public, and a list that reads them all alike hides that.

The split is on results rather than on authorship, and the two do not line up. The coal-alone test is in the first list and is not our method: we did not devise it, we have now run it, and both of those are true at once. Each entry says which it is.

Running a test does not promote the method. These protocols are ours whether or not we have answered them, none is validated against an instrument, and a result does not do that. What changes when a result exists is only that there is a figure of ours to argue with.

With a result of ours published

Four. In each the result is our own measurement with its conditions stated: how many repeats, what was and was not controlled, and what the figure does not cover. In three of the four the method is ours as well; in the fourth it is not, and running someone else’s test does not make it ours.

• The time-to-ash test. Minutes from ignition to full white-ash coverage, which our measurement makes a usage parameter rather than a quality one: it tells your customer how long to wait before the chip goes on, and it differs by format. Our figure for each format, the correction to the reading we first published, and what would falsify it.
• The dry-box test. Whether the moisture splitting your cubes was picked up in your warehouse or came with them. The protocol, both counts, and what the split means. On the cubes we ran from another producer the untreated three cracked and the dry-boxed three did not, which is the outcome that says cracking stock can be worth rescuing rather than written off.
• The coal-alone test — not our method, but now our result. Light one piece with nothing on it. If the coal alone is clean, the fault is in your blend and you can stop investigating the supplier. How to run it and how to read it. We ran it on one disc of ours and one instant-light disc bought on the local market and published what each did — both observations, and the step they deliberately do not take. One nose, one room, no instrument: it is an observation rather than a panel, and it establishes that a coal smells rather than what that does to a fragrance on top of it.
• The weight-gain test. How fast a cube takes on water in your market rather than in a laboratory. The protocol, and why we publish no shelf-life figure. Our one figure is for a sealed bag rather than a loose cube, which makes it a packaging result and puts it on the page where the foil decision is made. The loose-cube measurement is outstanding and nothing borrows the sealed one in its place.

No result of ours behind it

Three. Two are ours and unanswered — we publish the protocol and no figure, which is the honest state rather than an oversight. The third was never ours to answer.

• The sealed-jar test — our proposed method. Three cubes in a clean glass jar overnight, then smell the headspace: it separates a smell already in the product from one produced by burning. The protocol, and what it cannot tell you.
• The still-air test — our proposed method. How far the ash travels from the burner over a full session, which is the property no specification sheet quotes. The protocol, and how to read the scatter.
• The snap test. Not ours — documented trade practice, and the one test on this page with no result of ours behind it that is not waiting on us. Break a cube and read its core: a dark centre means it was never carbonised through, which is what sparks. How to run it and how to read it.

§4 · What we will not diagnose remotely

We do not assess another supplier's product from a photograph or a written description. A photograph shows colour and shape, the two properties least likely to explain why a cube smells wrong, and a description carries the assumptions of whoever wrote it. Run the tests above on your own stock and send us the figures instead, or run ours beside yours — the sample request page explains how.

Reviewed by Mohamed Noor, Director — 23 September 2026

§5 · What supports the triage above