What a COA is
A certificate of analysis is a laboratory report on a specific batch of product. It states what was tested, by what method, what result came back, and whether that result met a predefined specification.
The important word is batch. A COA is not a statement about a product line; it is a statement about the particular run of material that produced the box in your hand. This is why lot numbers matter, and why a certificate that names no lot is not really a certificate.
The anatomy of the document
- Product & lot
- What was tested and which specific batch. The lot should match the code printed on your box.
- Dates
- Manufacture, test and best-by. A test date long after manufacture, or a certificate with no dates, is worth asking about.
- Method
- The analytical technique — HPLC, ICP-MS, plate count and so on. Methods have known accuracy; naming them is a mark of a real report.
- Specification
- The range the result was required to fall within. Set before testing.
- Result
- What the instrument actually measured.
- Disposition
- Pass or fail against the specification.
Specification and result are the pair that carries the meaning. A result with no specification beside it cannot pass or fail anything — it is a number without a standard.
The three panels that matter
Identity
Confirms the ingredient is what it claims to be. This sounds trivial and is not: botanical adulteration and substitution are long-running problems in the supplement supply chain, and an identity test is the check against receiving something cheaper than what was ordered.
Potency (assay)
Measures how much of the active is actually present, usually as milligrams per serving against the label claim. Expect a tolerance band rather than an exact match — analytical methods have inherent variation, and a specification of, say, 90 to 120 percent of label claim is normal and honest.
Purity and contaminants
Three groups usually appear here:
- Heavy metals — lead, arsenic, cadmium and mercury. These occur in soil and water, so they are not evidence of wrongdoing; the point is that they are measured and held below defined limits.
- Microbial — total counts plus specific organisms such as Salmonella and E. coli, typically reported as absent in a stated sample size.
- Mycotoxins — mould-derived toxins, relevant to botanical and grain-derived ingredients.
Reading "not detected"
Two abbreviations cause more confusion than anything else on a COA.
LOD is the limit of detection — the smallest amount the instrument can see at all. LOQ is the limit of quantitation — the smallest amount it can put an accurate number on.
So "not detected" does not mean zero. It means below the limit of detection for that method. A sensitive method reporting "not detected" is a much stronger result than an insensitive one reporting the same words, which is exactly why the method should be named. A good report states the limit alongside the result.
"Not detected" is only as meaningful as the detection limit behind it. If the report does not state the limit, the phrase is not telling you much.
Who ran the test
An independent third-party laboratory has no commercial stake in the result. An in-house quality lab is a legitimate part of manufacturing, but it is the company checking its own work.
Look for the laboratory's name on the report, and ideally its accreditation — ISO/IEC 17025 is the relevant standard for testing labs. A report that names no laboratory is difficult to take seriously.
Six red flags
- No lot number. The certificate cannot be tied to anything you can buy.
- No dates. You cannot tell whether it describes current production.
- "Typical values" instead of measured results. That is a specification sheet, not a test report.
- Results with no specification. Numbers with nothing to pass or fail against.
- Only one COA on the whole site. Suggests one batch was tested, once, some time ago.
- The certificate is only available on request, slowly. Reluctance is itself an answer.
What good looks like
The standard worth holding a company to: you can find the lot number on your box, look it up, and read the report for that specific batch — including batches that have since expired, which stay published rather than quietly disappearing.
That last detail is the tell. A library that only contains current stock has been curated. One that keeps its history has not.
Ours is at the certificate library, indexed by lot and kept online for the life of the product.
Frequently asked
No, and a report claiming zero deserves more scepticism than one reporting a small number. Lead, arsenic, cadmium and mercury are present in soil and water everywhere, so trace amounts appear in agricultural products. What matters is that they are measured against a defined limit and fall below it.
It is the tolerance band for potency. Analytical methods have inherent variation and actives can degrade slightly over shelf life, so specifications are written as ranges. A result inside the band is a pass. A specification with no upper bound, or no band at all, is less rigorous than it looks.
Falsified certificates do exist. Protections against them: the report names an accredited independent laboratory, results are batch-specific rather than identical across lots, and the lab can confirm a report if contacted. Consistency in both directions is reassuring — real testing produces slight variation between batches, not identical numbers every time.
Usually cost. Per-lot third-party testing is a real recurring expense, and for a low-margin product it is the easiest line to cut. Occasionally it is because the results would be inconvenient. You cannot tell which from the outside, which is rather the point.
Not sure it is right for your pet?
Our eligibility screener checks your animal's medications and conditions against every ingredient on the formulation sheet. One minute, eight questions, and it will tell you if we are not a fit.