A certificate of analysis is a one- or two-page PDF laid out as a table. It is the most useful artefact a supplement brand can publish and the least widely read — partly because it looks forbidding, mostly because nobody explains what its rows are for. This page takes one apart section by section: what each part asserts, what it does not, and which absences matter.
- 0Sections a certificate is legally required to contain — the format is a convention
- 6Sections a complete one carries, from the header down to residual solvents
- 1Field that decides whether it describes the bottle in hand: the lot number
A companion piece asks who did the testing, and whether that claim can be checked at all. This one assumes the document has been opened. A report from an impeccable laboratory can still say nothing useful about the product somebody is holding.
The header decides whether the rest of the page is about this product or about a category
A lot-specific certificate carries at minimum: the material name, a lot or batch number, a manufacture date, the date the laboratory received the sample, the report date, the laboratory's name and address, a report number, and an authorisation.
The document that gets confused with it is a specification sheet — sometimes headed "typical analysis". Same table, same fonts, often the same header block. It describes what the material is supposed to be rather than what one batch measured, and it has no lot number because it is not about a lot. Legitimate, and not a test result.
One check costs thirty seconds: the lot number on the certificate should match the one stamped on the package. A certificate published as a permanent link also ages — the lot sells through, and the link keeps pointing at an unbuyable batch.
Identity answers a question most buyers assume is not in doubt
The identity section states that the material tested is the substance named. It is the least glamorous row on the page and the one the regulation is most insistent about: 21 CFR part 111, the US current Good Manufacturing Practice rule for dietary supplements, requires at least one appropriate test or examination to verify the identity of each incoming dietary ingredient, and has done since 2007.
What varies is how. Thin-layer chromatography and HPLC retention matching compare a sample against a reference standard; infrared spectroscopy compares a spectral fingerprint; DNA-based methods identify an organism, which matters when a species has close commercial relatives. The weak version of the row reads "Identity — Conforms" and nothing else. "Conforms" is a verdict; a method and a reference standard are an argument, and only one of them could be repeated by a second laboratory.
That pairing is why certificates are interesting at all. Testing is mandatory; publishing is voluntary.
The potency row is where the money is, and where the units do the deceiving
Potency — often headed "assay" — states how much of the active the material contains. Four things belong in the row: specification, result, unit, method. The failure mode is not falsification. It is that the number can be true about the wrong thing, in three routine ways.
The standardised-material trap. Many raw materials are sold as a percentage-standardised powder, so a "300 mg" figure may be 300 mg of a 50% standardised material — 150 mg of the named active. This desk has been caught by it: the alpha-GPC ranking carries a dated, public correction for exactly this error. A certificate usually resolves it, because a laboratory assays the active and reports it as the active.
The assay-target trap. On the mushroom shelf a crude polysaccharide assay and a specific beta-glucan assay return very different numbers from the same jar, and only one is the compound the research is about — the argument in the beta-glucan number. A potency row is only as meaningful as the analyte named in it.
The overage trap, in reverse. Manufacturers commonly fill slightly above label claim so the amount holds through shelf life, so 108% of label claim is ordinary rather than alarming. The result worth noticing is one near the bottom of the range — a 92%, say — passing with nothing to spare. That is what the specification column is for: "90.0–120.0% of label claim" states what counts as a pass before the result is read.
Contaminant sections are read by their units and detection limits before their results
Heavy metals. Four elements appear on essentially every supplement certificate — lead, arsenic, cadmium and mercury — generally run by inductively coupled plasma mass spectrometry against USP General Chapter <2232>, Elemental Contaminants in Dietary Supplements. Check the unit on the limit before the result: a limit per gram of powder and a limit per maximum daily serving are different quantities, converted by the serving weight. California's Proposition 65 adds a separate, stricter daily-intake trigger for lead that decides whether a warning appears on the package rather than whether the product may be sold — so a product can carry a California warning and sit inside the pharmacopoeial limit.
Microbials. Usually five rows: total aerobic microbial count, total combined yeasts and moulds, and the absence of Escherichia coli, Salmonella and Staphylococcus aureus. The counts are colony-forming units per gram against a ceiling; the absence rows are pass or fail against a stated sample weight, and that weight is part of the claim. USP chapters <2021> and <2022> are the methods most often named.
Residual solvents. Relevant only to extracted ingredients. It reports traces of the solvents used to pull a compound out of its source, classified and limited under USP <467>.
Then the trap running through all three. "ND" is not zero. It means not detected — by that method, at that method's threshold. A well-formed report prints the limit of detection or quantitation beside the result. A report printing "ND" alone says an instrument found nothing without saying how hard it looked.
Accreditation appears on the document as a scope, and the scope is the part worth reading
ISO/IEC 17025 accreditation is a statement about the laboratory rather than the product, and the companion piece makes that argument at length. Worth adding here is how it shows up on a certificate, because it is not a logo in the corner.
An accreditation comes with a scope: a published list of the specific methods, on specific material types, the laboratory has been assessed as competent to run. A lab accredited for elemental analysis of botanical powders is not thereby accredited for microbial enumeration, and a well-run one marks out-of-scope results with a footnote or an asterisk. Those footnotes are the most informative small print on the page, and they exist because the laboratory chose to be honest about a boundary. The related tell is the method column: a row naming a USP chapter or an AOAC method has connected itself to a published protocol, and "in-house method" has not.
A missing section usually means a test was not commissioned, which is a fact rather than an accusation
Laboratories are commissioned test by test, and each test is a priced line item. A certificate covering identity and heavy metals but not microbials generally means somebody paid for two panels and not a third — often because the third was run on the raw material at intake. Partial certificates are the majority, so the useful habit is not to score absences but to name the question each one leaves open.
- No lot numberNot about any particular batch. Everything on it may be accurate, and none of it can be matched to a product on a shelf. The one absence that empties the page.
- No method columnThe results cannot be reproduced or compared against another laboratory's. The finding may be sound; it is not checkable, which is the property the document exists to provide.
- No identity section on a botanical or fungal materialPotency assumes the material is what it is called, and for species with close commercial relatives that assumption is under the most pressure.
- No detection limits beside "ND"Reduces a contaminant panel to a set of assurances. A number, even a small one, carries more than a threshold-free "not detected".
A five-minute read of any certificate
Start at the top. Is there a lot number, and does it match the package? If not, stop — the rest of the document describes something else.
Then read the analyte name in the potency row before the number: standardised material or named active. That one word decides what the figure means. Then check whether the contaminant limits are per gram or per daily serving, and look for detection limits beside any "not detected". Then the method column. Finally, name what each missing section leaves open.
The cost per lot is also why so few get published. A published, lot-matched certificate is a strong signal precisely because it is expensive and optional; its absence is a weak one, because it is the default.
What this page settles
Four things. A certificate without a lot number describes a specification rather than a batch, and no other section repairs that. A potency figure means whatever its analyte says it means. Contaminant sections are read through their units and detection limits first, because "not detected" is a statement about a method rather than about a product. And an absent section usually means an uncommissioned test — an open question worth naming, not concealment worth alleging.
What the document cannot do is say whether the product is worth taking. Whether that quantity of that compound changes anything measurable is a question for trials, answered elsewhere on this site — in what the choline research measured.
The compendial chapters and the regulation cited here are named in the text. Standards are revised; the issuing body's own text is the authority, and anyone relying on a limit should read it there rather than take a summary on trust.