A Certificate of Analysis is the most information-dense document in the cannabis supply chain, and most of the people who receive one read three numbers on it. Total THC, total terpenes, and whether the contaminant page says PASS. The rest — a dozen minor cannabinoids, forty or more terpenes, the lab's own calculation method, the date it was issued — goes unread, because reading it properly takes an afternoon and a background in analytical chemistry.
The COA Report tool, now live on Cannavec, does that reading. You upload the certificate as a PDF; a few minutes later you get back an eleven-page report that works through the whole document and says, in plain terms, what it shows and what it does not. A sample report is linked at the end of this post, so you can judge the output for yourself.
What it does
The tool reads the certificate itself — not a form you fill in about it. It works out which kind of certificate it has been given (cannabinoids only; cannabinoids with a full contaminant panel; or the full picture with terpenes as well), and it handles certificates from any laboratory in any language. That detection matters, because a certificate with no terpene panel cannot support half the analysis a full one can, and the report is built to say so rather than fill the gap.
From there it produces eight sections:
| # | Section | What it answers |
|---|---|---|
| 01 | Cannabinoids | What the flower is: chemotype, THC:CBD ratio, and whether the lab's arithmetic checks out |
| 02 | Terpenes | What the profile is likely to feel like, and why |
| 03 | Chemovar fingerprint | How this batch compares with other batches sold under the same name |
| 04 | Indicated profile | Which conditions the evidence associates with this chemistry, graded A/B/C |
| 05 | Clinic guidance | Prescriber context, with numbered citations |
| 06 | Market position | Comparable UK products, normalised to £/g |
| 07 | Price recommendation | A floor, target and ceiling, with the reasoning shown |
| 08 | Contaminants | Each result checked against its stated limit, and what the panel does not cover |
Every clinical claim in sections 04 and 05 traces back to the Cannabis Knowledge Base and carries an evidence grade, so a reader can see whether a statement rests on randomised trials or on preclinical work. The report is written for professional and industry use; it describes associations in the literature, not treatments.
The arithmetic check most people skip
One small thing the report does on every certificate is worth drawing out, because it illustrates the approach. Labs report Total THC as a calculated figure: the small amount of active THC in raw flower, plus the THCA precursor multiplied by 0.877 to account for the mass lost when it converts on heating. The report recomputes that figure from the certificate's own raw values and checks it against the lab's stated total.
On the sample certificate, both Total THC and Total CBD reproduce to the thousandth of a percent. That is what a clean result looks like. When they do not reproduce — because a lab uses a different formula, applies a moisture correction, or has simply made an error — the report says so, and that discrepancy becomes a finding in its own right rather than something the reader has to notice.
Benchmarking against the archive
Section 03 is the part that did not exist before. Cannavec holds an archive of around 2,900 strain fingerprints, each a population-averaged terpene profile built from many independently tested batches and graded by how many separate laboratories and growers contributed to it. We wrote up how the archive was built, and why lab and supplier diversity matter more than raw sample count, in an earlier post.
The report uses that archive in two separate ways. First, a name lookup: it finds the archival fingerprint for the strain on the certificate and computes how closely this batch's terpene proportions match it. Second, a blind search: it takes the batch's terpene chemistry with the strain name withheld and asks which archived strains it most resembles. If the blind search returns something other than the named strain, that is reported.
For the sample, the name lookup found a Gold-tier Wedding Cake fingerprint built from 1,247 samples across 10 laboratories and 34 suppliers, and this batch matched it at 82%. The report is careful about what that number means. Cannabis is a natural product; terpene profiles shift with genetics, phenotype, harvest timing and growing conditions, and an 82% match on a fingerprint that deep is well within normal batch variation. The blind search agreed: nothing in the archive matched this chemistry better than Wedding Cake itself. Had the numbers come out differently — a low match, or a blind search pointing elsewhere — the report would say so in the same measured terms, because divergence is a finding about variation, not a verdict on the label.
Where it goes next
The report currently reads the same for everyone. That is about to change. A cultivator deciding how to price a batch, a pharmacy deciding whether to stock it, and a prescriber deciding whether it suits a patient are all reading the same certificate for different reasons. We are building role-specific versions: the same underlying analysis, same knowledge base, same fingerprint comparison, but with the sections ordered and weighted for what that reader needs to decide. The pharmacy version leads with the contaminant panel and market position; the clinical version leads with chemotype and the evidence grades; the cultivator version leads with the fingerprint comparison and price.
The archive behind section 03 is also growing as more public COA data becomes available, which will widen the geographic base of the fingerprints and improve coverage of strains that currently sit in the lower confidence tiers.
The sample report
Below is the full report generated from a real Full-panel certificate issued by a Quebec laboratory. The figures are read from the certificate; the readings beside them are written by the analysis.
Download the sample report (PDF, 11 pages)
The COA Report tool is available to Cannavec account holders at cannavec.ai/skills/coa. The same analysis can also be run inside a Claude session through the Cannavec MCP connector, for teams who would rather work from their own tools.