A COA (certificate of analysis) reads in three parts. First the header: which laboratory, which report number, which batch, which date. Then the cannabinoid table, with one line you should never skip: total THC. Finally the contaminants, if they were tested for — which isn't always the case.
If the batch number on the report doesn't match the one on your bag, you can stop there: the document describes a different product from yours.
A COA says nothing about taste or drying. It answers two questions: does the product contain what the label claims, and does it stay under the 0.3% THC threshold that determines whether it's legal in France?
Key points — A COA is a laboratory report that applies to a single batch: first check that the batch number matches your product. In the cannabinoid table, read the total CBD and total THC (neutral form + 0.877 × acidic form), not just the "CBD" and "Δ9-THC" lines. "ND" and "
What a COA is, and what it doesn't prove
A COA is a test report. A laboratory receives a sample, applies a method, returns results. Almost every report carries a line along the lines of "the results relate only to the sample submitted for testing." That matters: the laboratory generally didn't take the sample itself. It analyses what it's sent.
The consequence: a COA doesn't prove your bag is identical to the sample. That link rests on the seller's traceability, in other words the batch number. That's the limit of the system.
The header: laboratory, report number, batch, dates
The laboratory and ISO/IEC 17025 accreditation
The ISO/IEC 17025 standard (current version: 2017) sets the competence requirements for testing laboratories. A laboratory doesn't declare itself accredited: it is accredited by a national body — COFRAC (the French accreditation body) in France, its equivalents elsewhere in Europe.
The point almost nobody checks: accreditation covers a specific scope, meaning a list of tests. A laboratory can be accredited for water analysis and not for measuring cannabinoids. A report issued under accreditation normally carries the accrediting body's mark and a number; each laboratory's scope can be looked up on that body's website.
The batch number and sample description
Look for the line "sample," "description," or "client reference." It should contain a product name and a batch number. A vague description ("CBD flower," "hash") with no batch makes the report useless to you: nothing ties it to a particular product.
The dates
A report generally distinguishes the date the sample was received, the date of analysis, and the date of issue. No regulation sets a "validity period." But a batch of flower doesn't stay on sale indefinitely: an old report presented for a product sold today raises the question of which batch is actually involved.
The cannabinoid table, line by line
The method: HPLC or GC
HPLC (high-performance liquid chromatography) analyses the sample without heating it: it measures the acidic and neutral forms separately. GC (gas chromatography) heats the sample in the injector, which converts the acidic forms: it therefore generally gives an already "total" result. If the report is HPLC and shows neither CBDA nor THCA, ask why.
Acidic forms and neutral forms
The fresh plant mainly produces CBDA and THCA. Heat converts them into CBD and THC: that's decarboxylation. A dried flower contains a mix of both forms.
Total CBD and total THC: the calculation to know
The formula is the same for both molecules:
- Total CBD = CBD + 0.877 × CBDA
- Total THC = Δ9-THC + 0.877 × THCA
The 0.877 coefficient corresponds to the mass the acidic molecule loses as it decarboxylates. Fictional example: a report shows CBD 2.1% and CBDA 17.0%. Total CBD works out to 2.1 + (0.877 × 17.0) = 17.0%. Reading only the "CBD" line would make you think the flower was 2%.
Same logic on the THC side, with a regulatory stake attached. Fictional example: Δ9-THC 0.05%, THCA 0.25%. Total THC works out to 0.05 + 0.22 = 0.27%. Showing only "Δ9-THC: 0.05%" gives an incomplete picture.
On the legal side, as of September 2026: the arrêté (ministerial order) of 30 December 2021 authorises products whose Δ9-THC content doesn't exceed 0.3%. The method used to control hemp in the field, for its part, measures total THC by construction. And the ANSM (the French medicines safety agency) added THCA to the list of narcotics in 2024, excluding it when the THC content stays under the 0.3% threshold. The cautious reading is therefore to require total THC under 0.3%. Check the current state of the law when you're reading this, for instance in our overview of the legal framework for CBD in France.
Minor cannabinoids: CBG, CBN, CBC
CBG and CBC generally appear at low levels in a CBD-dominant flower. CBN is a special case: it comes from THC breaking down over time, with light and oxygen. A clearly detectable CBN level on a flower is therefore more a sign of ageing than a selling point. To place these molecules relative to each other, see our guide CBD, CBG, CBN, CBC.
Units
Results are expressed as % by mass or in mg/g: 10 mg/g = 1%. For flower, check whether the result is given "on a dry-matter basis" or "on the product as sold": the same sample shows a higher percentage once the water is stripped out of the calculation.
LOQ, LOD, ND: reading the "empty" boxes
- LOD (limit of detection): the smallest quantity the method can distinguish from background noise.
- LOQ (limit of quantification): the smallest quantity the method can measure reliably. It is always higher than the LOD.
- ND (not detected): below the LOD.
: possibly detected, but too low to be measured.
None of these mean "zero." It all depends on the threshold. A THC reading of "ND" with a LOQ of 0.01% is solid information. The same "ND" with a LOQ of 0.2% says almost nothing against a legal threshold of 0.3%.
Another line worth looking for: measurement uncertainty, often marked "±". No measurement is exact to the decimal point. A total THC result very close to 0.3% doesn't carry the same weight as a result clearly below it.
Contaminants: only if that testing was ordered
A "cannabinoids only" COA is common: each family of contaminants is a separate test, billed separately. That's not fraudulent, but such a report tells you nothing about how clean the product is.
- Heavy metals: lead, cadmium, arsenic, mercury. Hemp readily absorbs metals from the soil — it's actually studied as a way to decontaminate land. The soil's origin matters.
- Pesticides: multi-residue screening, covering anywhere from dozens to several hundred molecules. Look at how many substances were screened and the LOQs, not just a "compliant" label.
- Microbiology: total microbial count, yeast and mould, E. coli, salmonella. This is the most relevant test for a dried flower that's been poorly stored.
- Mycotoxins: aflatoxins (B1, B2, G1, G2) and ochratoxin A, produced by certain moulds.
- Residual solvents: relevant for extracts made using a solvent. A flower, a dry-sieved hash, or an ice-water extraction aren't affected.
We found no French regulation setting limits specific to CBD flower for these contaminants: laboratories compare results against existing reference frameworks (food, plants, pharmacopoeia). If the "limit" column doesn't state its source, ask for it. And a reminder: a flawless COA doesn't make combustion harmless. Smoking any plant material is harmful to your health.
Matching the COA to the label
Put the product sheet and the report side by side, and check four points.
- The batch: identical on the bag and on the report.
- The CBD level: the product sheet should match the total CBD figure. A flower listed at 19.45%, like the Tyson, or a resin listed at 44.96%, like the Candy Cream, should line up with a report giving a total CBD in the same range, allowing for uncertainty.
- The category: a flower report doesn't apply to a resin, even from the same variety. Resins and hash concentrate the trichomes: every cannabinoid rises, THC included. This is where the total THC line deserves the most attention.
- Overall consistency: in the plant, CBD and THC are produced together. A very high CBD level paired with a total THC of "ND" is worth a question about the LOQ used.
These habits apply to any seller, in a shop or online. A seller who can't provide a report tied to the batch leaves you with no way to check what you're buying.
The most common reading mistakes
- Reading the "CBD" line instead of total CBD, and concluding the product is under-dosed.
- Reading "Δ9-THC" alone and ignoring THCA.
- Taking "ND" to mean "0" without checking the LOQ.
- Confusing mg/g and % (a factor of 10).
- Believing that a COA with no contaminants section proves there are no contaminants.
- Trusting the ISO logo without checking the scope of accreditation.
- Accepting a report whose batch doesn't match, "because it's the same variety."
If the document itself looks doubtful — a patched-together layout, suspiciously round numbers, an untraceable laboratory — the logical next step is our guide to spotting a fake COA.
Sources
- Arrêté du 30 décembre 2021 portant application de l'article R. 5132-86 du Code de la santé publique (seuil de 0,3 % de Δ9-THC) — legifrance.gouv.fr
- ANSM, « L'ANSM inscrit de nouveaux cannabinoïdes sur la liste des stupéfiants », décision du 22 mai 2024, applicable le 3 juin 2024 — ansm.sante.fr
- Norme ISO/IEC 17025:2017, « Exigences générales concernant la compétence des laboratoires d'étalonnages et d'essais », ISO.
- COFRAC (Comité français d'accréditation) — recherche des organismes accrédités et de leur portée.
- Pharmacopée européenne (EDQM, Conseil de l'Europe) — référentiel pour les contaminants des drogues végétales.



