An ISO/IEC 17025 accredited laboratory is one whose competence at named test methods has been assessed by a national accreditation body, and is reassessed on a cycle. For coffee it takes a representative sample of the green bean, extracts it, and screens it by LC-MS/MS for specific compounds. It reports against a limit of detection, never against zero.
The standard itself is dry reading. What the laboratory does is simple enough to say in plain words.
What the accreditation covers
It covers the laboratory, not the coffee. ISO/IEC 17025 is titled general requirements for the competence of testing and calibration laboratories, and that is what an assessor examines: staff competence, calibrated equipment, validated methods, traceable reference materials, how measurement uncertainty is calculated, how records are kept, how a wrong result would be caught before it left the building.
Accreditation is granted scope by scope. A laboratory can be accredited for ochratoxin A by LC-MS/MS and not accredited for something else on the same bench. So the useful question is not whether a laboratory is accredited, but whether the method on your report sits inside its scope. Most national accreditation bodies also belong to a mutual recognition arrangement, which is why a report issued in one country is accepted in another.
Sampling is the hard part
Mycotoxin contamination is not spread evenly through a lot. It concentrates in a small number of beans, sometimes a very small number, which means a single handful proves close to nothing. Careful practice takes many small increments from across the whole lot, combines them into one aggregate sample, mills that finely, and draws the test portion from the milled material.
In the published work on mycotoxin testing, sampling error is usually the largest single source of uncertainty in the exercise, larger than the analysis itself. A very good laboratory handed a bad sample will produce a precise answer to the wrong question.
LC-MS/MS, in plain words
Liquid chromatography with tandem mass spectrometry. Three steps, one after the other.
Extraction, first. The milled coffee is shaken with a solvent that pulls the target compounds out, and the extract is cleaned up so that coffee's own chemistry does not crowd the instrument.
Separation, second. The extract is pushed through a column, and compounds emerge at different times according to how strongly they cling to it. Ochratoxin A leaves the column at its own moment, and that moment is part of its identity.
Weighing, third. The mass spectrometer selects a molecule of the expected mass, breaks it apart, and weighs the fragments as well. Two weighings of the same molecule, parent and fragment, is why the identification is specific rather than a good guess. Another compound of similar mass will not shatter into the same pieces.
Limits of detection, and the phrase not detected
Every method has a floor. Below it, the signal cannot be told apart from background noise. The limit of detection is the lowest concentration a method can see at all. The limit of quantification, a little higher, is the lowest it can put a number on.
So a result reading not detected means below the limit of detection for that method, on that sample. It does not mean zero. Nothing can mean zero, which is exactly why the phrasing matters. A laboratory reporting not detected against a stated limit has told you something exact. A brand promising toxin free coffee has told you something no laboratory would sign.
Our published results read not detected, against the laboratory limit of detection, for ochratoxin A, aflatoxins B1, B2, G1 and G2, and pesticide residues. Three coffees are covered. The others carry no report and are not scheduled for one; their pages say so. The ledger is here.
Independence, and what we send
Independence means the party that paid for the coffee is not the party judging it. A roaster grading its own homework is not a test, it is a claim. What makes a third-party report checkable is dull detail: a report number, a date, the method named, the accreditation mark of the body that assessed the laboratory, and the compound list with its limits.
For the coffees we test, we send green beans, before roasting. Two reasons. Roasting is a variable, and a green-bean result describes what actually arrived in Malaysia. And a tested lot that does not clear never becomes coffee, which is the only version of testing that changes anything. Every lot we buy is one farm and one harvest, which is what makes a sample worth taking at all; the six coffees are here, and the wider standard in our standard. The compound list is set out in what we test for, and why.
The short answers
What does ISO 17025 accreditation mean?
That a national accreditation body has assessed a laboratory's competence at specific test methods, and reassesses it on a cycle. It describes the laboratory and its methods, not the quality of the coffee tested.
Does not detected mean zero?
No. It means the compound sat below the limit of detection for that method, on that sample. No test can report zero, so not detected against a stated limit is the honest phrasing.
Why test green coffee rather than roasted coffee?
Because a green-bean result shows what arrived, before roasting adds a variable, and because a lot that does not clear can still be turned away rather than sold.
Can a roaster test its own coffee?
It can measure things, but it cannot verify itself. An independent accredited laboratory has no stake in the result, and its report carries a number, a date, a named method and its limits of detection.
