Third-party testing is the industry's standard — but what does one Certificate of Analysis actually cover? A look at what testing measures, what it doesn't, and the questions around publicly available test data that researchers should understand.
When a research peptide supplier advertises "third-party tested" or links to a Certificate of Analysis (CoA), it signals something meaningful: someone outside the company analyzed a sample and reported the results. In an industry with no regulatory oversight, this is the closest thing to quality verification that exists.
But the signal is weaker than it appears. Understanding the gap between what a test proves and what vendors imply it proves is essential for anyone evaluating suppliers in this space.
A standard GLP-1 peptide test from a lab like Janoshik Analytical — the most widely used testing service in the research peptide space — typically includes:
Janoshik's GLP-1 blind test — their most popular product for this space — covers all three of these checks for semaglutide, tirzepatide, and retatrutide. It costs $300, takes roughly 5 days, and results are returned to whoever submitted the sample.
A CoA certifies the result for the specific sample that was submitted. It says nothing about:
A supplier might test one batch every few months and use that single CoA to represent months of inventory across multiple production runs. Unless the CoA shows a lot number that matches your specific order, the document is at best circumstantial evidence of what you received.
This is not unique to peptides — pharmaceutical manufacturers face the same issue and address it through lot-level testing, regulatory oversight, and chain-of-custody documentation. The research peptide market has none of those structures.
Janoshik maintains a public database of test results that anyone can browse. Scroll through it and you'll notice that purity numbers are almost uniformly high — 99.1%, 99.7%, 99.8%. Failure results are rare or absent.
This raises a structural question worth understanding.
Publishing a result to the public database is optional and costs an additional $70 on top of the $300+ test fee. Janoshik is a legitimate, independent testing laboratory — the selectivity question is not about their practices, but about which results vendors choose to submit for public posting. A vendor who receives a passing result has an obvious incentive to pay the publication fee; one who receives a failing result may not. As one community resource, Peptide Protocol Wiki, has observed:
"Vendors do not pay $70 to broadcast failures." — Peptide Protocol Wiki
Whether this economic incentive meaningfully shapes the database's composition is an open question — it's possible that vendors with consistent quality genuinely do produce clean results at a high rate. But it's a question worth holding when reading the database: does it represent a random sample of all tests run, or a vendor-curated subset? The Janoshik public page itself notes they have no affiliation with the companies listed — they provide the analytical service; what vendors do with results is up to the vendor.
You can see an example of a published passing result here: Semaglutide 10mg — Test #93542. Results are delivered as a scanned lab report image, not machine-readable data, which makes systematic analysis of the full dataset difficult.
When testing is conducted without vendor involvement — by community members, researchers, or journalists submitting purchased samples without the vendor's knowledge — published figures tell a different story than the public database. It's worth understanding what that data says, and also its limitations.
These figures come from community-compiled testing campaigns where the samples were not vendor-selected. They are widely cited in research peptide communities (Peptide Protocol Wiki, Peptide Grades, and related forums) but should be understood as community-reported data, not peer-reviewed science. Independent testing campaigns also carry their own potential biases — buyers who suspect a problem are more likely to test, which could skew failure rates upward relative to the true market average. The actual market-wide failure rate is not known.
That said, if these figures are roughly accurate, they suggest a meaningful gap between vendor-submitted test results and unannounced independent sampling. In at least one community-reported case, a semaglutide sample from a lower-tier vendor had a molecular weight consistent with a different compound entirely — not an impure semaglutide sample, but apparently a different molecule. Whether this is representative or an outlier is genuinely unclear.
The contamination question — compounds like exendin-4 analogs appearing in retatrutide samples — is particularly worth noting because it falls outside what a standard purity percentage captures. A product could pass an HPLC purity test and still contain biologically active contaminants if they are not specifically screened for.
This is an open question, and an important one.
When a vendor tests a batch and it fails — say, 84% purity instead of 99% — what happens to that inventory? Is it discarded? Sold anyway without the CoA reference? Blended? There is no public record, no regulatory requirement to report, and no audit trail. The research peptide market has no mechanism to answer this question.
There are a few plausible scenarios. Some vendors with strong manufacturing standards may genuinely discard failing batches — the CoA serves as a real quality gate. Others may sell the inventory without prominently linking it to any test document, since most buyers never ask for lot-specific CoAs. Some may not test routinely at all and submit only when they expect a passing result.
An alternative, more charitable reading: the peptide synthesis process — at least for the established suppliers — may be sufficiently controlled that failures are genuinely rare in their production, and the 43% failure rate is concentrated in lower-tier or overseas-sourced products. This is possible. The data does not clearly separate vendor tiers in a way that resolves the question.
The cost structure of testing matters for understanding how much of it happens. A basic GLP-1 blind test from Janoshik:
| Cost Item | Amount |
|---|---|
| GLP-1 blind test (identity, purity, amount) | $300 |
| Public database publication (optional) | $70 |
| US domestic shipping to Prague lab | $50–75 |
| Rush processing surcharge (if needed) | +100% of base |
| Total for one published result | ~$420–445 |
For a small vendor selling 5mg semaglutide at $80 per vial, a single published CoA costs the equivalent of 5–6 vials sold. If they test quarterly and publish, that is $1,700+ per year just for documentation on one compound. A vendor with 10 peptides in their catalog faces real economic pressure on testing frequency.
This does not mean testing is not worth doing. It means that the economics favor larger, more established suppliers who can spread testing costs across higher volume — and favor testing the flagship GLP-1 compounds (semaglutide, tirzepatide, retatrutide) over the full catalog.
Given these limitations, what should a researcher look for when evaluating a supplier's test documentation?
Several community resources have emerged to aggregate test data and vendor reputation beyond the public Janoshik database. Finnrick.com maintains independent vendor ratings based on testing history, community reports, and product documentation quality. BetterNewLives.com links to Finnrick vendor ratings in our price comparison tables where available — identified as a "🔬 rating" link next to supplier names.
These aggregated ratings are a useful complement to individual CoAs because they reflect patterns across multiple tests and batches over time, rather than a single data point chosen by the vendor.
Testing matters, and there are real differences between suppliers in how seriously they approach it. A vendor who regularly submits product to independent labs, publishes results with specific lot numbers, and links CoAs to particular batches is doing something substantively more than one who posts a single generic document. That distinction is worth recognizing.
At the same time, a single published test result — even a perfect one — leaves several questions unanswered:
These are genuinely open questions — not accusations. Some vendors have robust quality processes where a CoA represents a real quality gate; others may use testing primarily for marketing purposes. The community-reported data, whatever its limitations, raises the possibility that the gap between these two groups is larger than the uniform appearance of the public database might suggest.
We don't have definitive answers. What we can say is that more information — community tests, lot-matched CoAs, independent aggregators like Finnrick — gives a more complete picture than any single vendor-supplied document can.
If community-reported failure rates in unannounced testing are roughly accurate, and if the public database is structurally weighted toward passing results, what does a single published CoA actually tell a researcher about the specific product they're receiving? The answer likely varies significantly by supplier — and that variability itself may be the most important thing to understand.
BetterNewLives.com tracks current prices for semaglutide, tirzepatide, retatrutide, and other research peptides across US suppliers — with third-party rating links where available.
View Price Comparison →