Third-party testing proves one narrow thing: an independent laboratory received a sample, ran a defined list of tests, and reported what it found. It does not certify a brand, a product line, or the next batch off the line. And it establishes nothing about any compound that was not on the analyte list, because a certificate cannot clear what it never looked for.
That last point is the one skipped most often, and the one that matters most. What follows is the distance between what a report establishes and what a shopper reads into "third-party tested" on a box.
What "third-party" actually means
A third-party test is one performed by a laboratory with no ownership, financial, or corporate relationship to the company selling the product. The conventional numbering clarifies the whole question:
- First party — the manufacturer testing its own goods. Internal quality control.
- Second party — the buyer testing goods before accepting them. A distributor commissioning its own panel.
- Third party — a laboratory with no stake in the answer, hired to report what it finds.
The word doing the work is independent, not external. Sending samples to a laboratory your parent company holds a stake in is external testing; it is not third-party testing. Neither is an arrangement where the laboratory is contractually limited to the analytes the client specifies and obliged to keep unfavourable results confidential. Those arrangements exist, they are legal, and nothing on the face of a certificate reveals one.
You assess independence from three fields: the laboratory's name and physical address, the customer of record, and the accreditation line. If lab and brand share an address or a parent, or the lab has no findable corporate footprint, "third-party tested" is decoration.
In-house QC is not worthless. It is just not evidence for you.
It would be wrong to treat internal testing as a red flag. Any serious manufacturer runs continuous in-house quality control: raw-material identity, moisture, weight variation, in-process assays, stability. Federal manufacturing rules for dietary supplements require companies to set specifications for identity, purity, strength, and composition and to verify them. A company doing that properly tests more often, and closer to the line, than any contract laboratory it hires.
The problem with internal QC is not competence. It is that you cannot audit it. The record lives inside the company, and so does the decision about what to do with an unfavourable result, and so does the commercial pressure bearing on that decision. None of it is visible to someone holding a package in a shop.
Why the distinction is commercial before it is scientific
An in-house chemist can be better at the bench than a contract lab analyst. The case for third-party testing was never about who is more skilled. It is about who decides, and who can check.
When a result comes back unfavourable, someone decides whether to retest, whether to release the lot anyway, and whether the original record is kept. In a first-party structure that decision sits with the party holding the inventory. A third-party structure moves the analysis out of that room. It does not move the decision about what to publish — brands still choose which certificates reach the website — but the document was produced by someone who did not know which answer was wanted.
Downstream, everyone treats the difference as decisive. Retailers, payment processors, insurers, and state regulators read a report from a named independent laboratory as diligence, and an internal record as an assertion. A third-party certificate is transferable evidence; an internal QC log is not. That is why we ask for one before a brand reaches our shelf, and the standard is written down in how we vet the brands we carry.
Batch-level testing versus the one-off report
The most common failure in this category is not fake testing. It is real testing, done once, cited forever.
Botanical raw material varies. Harvest region, season, supplier, drying conditions, and extraction run all move the numbers. Contaminant risk is a property of a specific lot, not of a brand. A heavy-metals pass on material bought in one quarter is not evidence about material bought two quarters later from a different broker.
So a certificate is worth as much as its lot linkage. Signals of a one-off rather than a programme:
- No lot number, or a lot field reading "N/A." The report cannot be tied to anything you can buy. Not hypothetical — we have read certificates in this category with the lot printed as N/A and a generic sample name. Batch-level in form only.
- A report date years earlier than the product in your hand. Dating is not a formality.
- One report standing in for a whole catalogue. Three documents behind fifty SKUs is a marketing asset, not a testing programme.
- Method groups dated "TBD." A panel printed with no analysis date has not been run.
What batch-level looks like instead: a unique lot identifier matching the code on your package, dates consistent with when that lot was made, one report per lot, and a library that keeps expanding. Everything we hold, gaps included, sits on our certificate of analysis page.
Chain of custody: the part nobody photographs
A certificate reports on the sample the laboratory received. Everything before that is chain of custody, the weakest link in botanical testing.
Who pulled the sample? From where in the lot? Was it random or hand-picked? A brand that hand-selects a sample and couriers it to a laboratory has tested that sample honestly. Whether it represents the pallet is a sampling question the certificate cannot answer.
The fields showing a chain of custody exists at all: a date received stated separately from date analysed and date reported; a laboratory-assigned sample ID distinct from the brand's lot code, plus the initials of whoever received it; and a note on condition on receipt. Some certificates print a photograph of the unit tested.
Most certificates also state that results reflect only the samples indicated, as received — a disclaimer marking where the laboratory's responsibility begins. The stronger arrangement, in which the laboratory or an independent agent draws the sample from the lot itself, is rare here. We would rather say so than imply a rigour the paperwork does not support.
What a certificate genuinely establishes
| What the certificate shows | What that establishes | What it does not establish |
|---|---|---|
| Named laboratory with an accreditation mark | An assessed facility did the work | That every method sat inside the accredited scope — unpacked in ISO/IEC 17025 explained |
| A lot number matching your package | The results apply to the units you hold | Anything about any other lot |
| "Not detected" beside an analyte | It was not found at or above the method's detection limit in that sample | Absolute zero, or anything about analytes not listed — see what non-detect means |
| An assay or potency figure | A measured quantity of one named constituent | Anything about what the product does |
| Contaminant panels within limits | Those contaminants fell under the stated specifications | Anything about contaminants outside the panel, or panels marked "not tested" |
| An authorised signature | The laboratory stands behind the report | That the brand's label is accurate |
Four things buyers assume that certificates do not say
- "Tested" means safe. A certificate is a measurement, not a verdict. Whether a product suits a particular adult is a question no laboratory report has an opinion on.
- "Clean" means clean of everything. Every panel is finite. A pass on a four-metal screen says nothing about a fifth metal, and a clean alkaloid screen says nothing about a compound that was never an analyte.
- A COA certifies the brand. It certifies a sample. Brands are not certifiable objects; batches are.
- An accredited laboratory produces accredited results. Accreditation is granted method by method. A lab can hold genuine accreditation and still print results from methods outside that scope, disclosed in a footnote most readers skip.
The panel is the ceiling, including ours
Our own position, without softening. The supplier certificates we hold cover, and report non-detect for, seven analytes: 7-hydroxymitragynine, mitragynine, speciociliatine, speciogynine, paynantheine, corynantheidine, and mitraphylline. That is what lets us say the products we carry contain no kratom, no mitragynine, and no 7-OH.
Those panels do not include mitragynine pseudoindoxyl, MGM-15, MGM-16, or corynoxine B. So we do not write "no MGM," we do not write "no pseudoindoxyl," and we will not write "free of all scheduled compounds" or "contains no Schedule I substances." Not because we think those compounds are present, but because the panel never looked, and a non-detect you did not run is not a result. Declining those sentences costs us copy competitors use freely.
The reason this is live rather than academic: DEA filed two Notices of Intent on July 1, 2026, published in the Federal Register on July 6, 2026. FR Doc 2026-13580 covers 7-hydroxymitragynine above a threshold of 0.050% by weight, or 1.00 mg per article. FR Doc 2026-13581 covers mitragynine pseudoindoxyl, MGM-15, and MGM-16 with no threshold at all. Each takes effect on or after August 5, 2026, on the date the order publishes; each runs two years, extendable by one; and neither is subject to judicial review under 21 U.S.C. 811(h)(6). There is no grace period and no sell-through window.
Every compound named in that second notice — mitragynine pseudoindoxyl, MGM-15, MGM-16 — is a compound our panels never looked for. That is why the gap gets published rather than papered over.
Four questions worth putting to a brand in writing
- Which laboratory, and what is its accreditation certificate number?
- Is testing performed on every production lot, or on a sampling schedule?
- What is the full analyte list, and which compounds are deliberately excluded?
- Are any methods on the certificate outside the laboratory's accredited scope?
A brand that answers all four in writing is unusual. A brand that answers none has told you something. For the mechanics of the document, our guide to reading a certificate of analysis covers the header block, the results columns, and the units.
Frequently Asked Questions
Is third-party testing legally required for supplements?
No. Federal manufacturing rules require companies to set and verify specifications, but they do not require an independent laboratory to do the verification, and they do not require anyone to publish a certificate. Third-party testing and public COA libraries are voluntary. That is why their presence is informative, and why their absence is too.
If the brand pays the laboratory, is the result still independent?
Someone has to pay, and in commercial testing it is nearly always the party submitting the sample. Independence comes from structure, not from who writes the cheque: no shared ownership, an accredited method, a named analyst, and a report issued under the laboratory's own name. What payment does control is the analyte list — which is why a panel's contents deserve as much scrutiny as its results.
Does a certificate say anything about how a product will affect me?
No. A potency figure describes what is in the sample, not what the sample does.
Can a certificate prove a product contains no kratom?
It can establish that the kratom alkaloids on the panel were not detected in that sample at or above the method's detection limit. On the certificates we hold that covers seven compounds — a substantive result, and the basis for the claim we make about our Cat's Claw collection. It cannot establish anything about compounds outside the panel.
What if a seller will not produce a lot-matched report?
Treat the refusal as the answer. Any distributor running a real COA library can retrieve a document by lot code in minutes; one that cannot has nothing to retrieve.
These statements have not been evaluated by the Food and Drug Administration. These products are not intended to diagnose, treat, cure, or prevent any disease. For adults 21+. Legality depends on the specific formulation and applicable state law — check the rules where you live. Nothing here is legal advice.