How we audit, step by step.

01 / Public lab records

Every score starts from lab records anyone can check. We keep dated mirrors of public third-party lab data: 1,828 certificates from the Kovera Labs verifier and 337 readable reports from Janoshik's public-tests database, plus the smaller Liquilabs and MZ Biolabs sets described below. Each mirrored record carries the date we retrieved it.

For each vendor we count the reports that name it, read the purity and identity results, and note the fails as well as the passes. A vendor with no record in any mirror is scored on its own published COAs alone, and the profile says so.

02 / COA cross-check

Vendors publish their own COAs. We check each one against the issuing lab's own record where the lab offers one: the certificate number, batch ID, compound, purity and date must match what the lab holds. A COA the lab cannot find, or one that differs from the lab's copy, is noted in the audit and counts against the COA subscore.

We also look for what a vendor leaves out. When a lab's public record shows a failed report for a batch the vendor presents only with a passing certificate, the profile lists both.

03 / Regulatory record, prices and service

We search FDA warning letters, federal court dockets and state regulator actions for every vendor, its company names and its owners. An open FDA warning letter takes 1.5 points off the composite. A documented state regulatory action, such as a state board of pharmacy order, takes 0.5 off. Selling a compound under a code name of the kind FDA has cited, with no letter, takes 0.5 off as regulatory exposure. Each deduction is stated on the vendor page.

Pricing is scored per milligram against the median price for the same compound and vial size across the vendors we track, at the sale price a buyer would pay on the day we read the shop. That date is printed with the price.

Shipping and support are scored from the vendor's published policies, its payment options, and the broader user-reported record: forum threads, review profiles and third-party raters, each cited. These are second-hand signals, and the profile says which ones we used.

Public review profiles, Trustpilot first, are recorded on the vendor page with the date we read them: the score, the number of reviews, the share of five-star and one-star reviews, the date of the oldest review, whether the company invites reviews, and the complaints that recur in the low-star reviews. A high score earns nothing on its own. Nearly every peptide shop on Trustpilot sits between 4.3 and 5.0, a shop can keep a perfect score after it closes, and invited reviews can arrive in bursts. What moves a score is the negative record (non-delivery, payment trouble, refused refunds, underfilled vials) and a review history built in weeks rather than years. A review profile never outweighs a lab result, and if the review site will not let us read the profile, the vendor page says it was not checked.

What we have not done yet

We have not placed a first-hand test order with any vendor. We have not bought a vial, logged a delivery, or sent a sample to a lab ourselves. Every score on this site is a public-data audit. First-hand orders are a planned step. When one becomes part of a vendor's score, that vendor's page will say so, with the order date and the lab report.

04 / The composite score

We don't average the four subscores — we weight them. Weighting reflects what actually matters for a research buyer:

• Purity / COA quality — 40%
• Pricing — 25%
• Shipping — 20%
• Customer service / support — 15%

A score of 90 or above means “buy with confidence.” 70–89 is “works, with caveats.” Below 70 means we've found enough red flags that the audit notes deserve a careful read before ordering.

05 / Refresh cadence

Each vendor page shows the date we last read its records. We re-read a vendor when its lab record changes, when FDA or a court files something new, when it changes lab partners, payment processors or shipping carriers, or when a reader sends a COA that contradicts our finding. Score changes are noted in the corrections log.

06 / What we don't test

We don't evaluate clinical safety, dose efficacy, or human-use outcomes. Compounds discussed on this site are referenced from peer-reviewed literature and sold by vendors as research-grade material.

How third-party peptide testing works

Independent peptide testing pairs reverse-phase HPLC with mass spectrometry to answer the two questions a buyer actually has: is this the compound on the label (identity), and how much of it is what the label says (purity). HPLC separates the sample by hydrophobicity and quantifies the main peak against impurity peaks to produce the headline purity percentage. Mass spec confirms the molecular weight matches the target sequence, which is how identity gets verified.

Janoshik Analytical runs this pipeline on the bulk of the publicly-visible peptide-testing corpus. Our local mirror of their public-tests database covers thousands of tests across 200+ compounds. The full purity analysis read is here.

We mirror three additional third-party labs to widen the picture. Liquilabspublishes Particle Peptides' COAs from the Czech Republic. MZ Biolabs runs Ascension's testing. Kovera Labs began publishing in early 2026 and serves a US-facing slice of the market. A lab is treated as a credible source when its corpus is broad across vendors and its failures are visible in public, the same criteria covered in our Janoshik audit.

Certificate references are not always permanent. Kovera re-certifies records in place under the same certificate number, so we store a retrieval date with every mirrored certificate and re-run the full corpus periodically. A refresh on 1-2 August 2026 found 57 of 1,828 records revised since we first recorded them, none of which moved a result across the 98 percent pass threshold. Every COA figure we publish is current as of its stated retrieval date.