How we work

Exactly where every number on the site comes from, how it's checked, and what it can't tell you.

Prices

• Source: each vendor's own website. A separate scraper per vendor reads its product listings: name, vial size, number of vials, price, currency and stock. 20,714 listings from 240 vendors; last collected October 7, 2026.
• Which vendors: every vendor that Finnrick has tested (465) is checked; those with a working online shop get prices.
• Price per mg = price in US dollars ÷ total milligrams (vial size × number of vials). Other currencies are converted at fixed reference exchange rates, so non-USD prices are approximate.
• Names are matched to one canonical compound (for example "Reta", "GLP-3 RT" and "Retatrutide" are the same), and blends are split into their components so the planner can use them.
• Sanity checks: listings far below or above the usual price per mg are marked "check vials" or "kit price?" and left out of the "cheapest" pick until confirmed. Shipping costs, minimum orders and how you order (checkout, account, message) come from each vendor's site.
• Default view: vendors scoring 80%+ on Finnrick, in stock, shipping to the US, sorted by price per mg. You can change every filter.

Vendor scores

Vendor scores, test counts and per-peptide ratings come from Finnrick, which buys vials anonymously and tests them for purity and amount. We show Finnrick's purity-focus score (99.5%+ purity and dose within ±35%). We don't run lab tests ourselves and can't vouch for a vendor beyond what testing shows: a high score means past samples tested well, not that every batch will.

The metascore

A 0–100 score we work out ourselves from the raw evidence, so no single source's verdict decides a vendor's rank:

• Quality of bought vials (65%): every test of a vial bought like a customer would (Finnrick's public results and approved submissions), scored on dose (full marks within 90–115% of label), purity (full marks at 99.5%+) and identity; a failed identity or endotoxin test scores zero. Newer tests count more (a year-old test counts half), and vendors with few tests are pulled toward the market average so two lucky results can't top the chart.
• Honest paperwork (20%): our review of the vendor's published lab reports, how many we confirmed on the lab's own site, PepIndex's transparency score where it has one, and a cap when results look too uniform to be real.
• Claims match purchases (15%): whether vials bought anonymously come in underdosed much more often than the vendor's own reports show. Left out when there's nothing to compare.

The confidence label (low, medium, high) reflects how many independent tests sit behind the score: under 3, under 10, or 10 and more. Finnrick's own score is still shown next to ours.

Vendor lab reports (COAs)

• We checked the websites of 240 vendors and collected 5,188 certificates of analysis that 188 of them publish, recording the lab, date, measured purity and measured amount versus the label.
• Numbers are taken only from text we can read (web pages, PDF text, lab verification pages). Image-only reports are listed without numbers, never guessed.
• Lab names are standardized, and only labs we can confirm as independent count as "third-party".
• Possible problems are flagged when reports are issued to another brand, look edited or generated, show the wrong compound, reuse one lot number across unrelated products, or carry future dates. Honest underdosing shows in the numbers instead.
• Every flag is then re-checked against the public evidence (the vendor's pages and the labs' own verification pages) and graded:
  • Severe: fabricated or borrowed reports. Reviewed for a ban.
  • Major (19 vendors): important reports belong to someone else, or one report is used as proof for different lots or compounds. Shown as a red "COA concerns" badge with the reason.
  • Minor (32): an isolated mislink, date error or supplier report among otherwise consistent reports. Shown as a grey "COA note".
  • Cleared (3): the finding had a legitimate explanation, such as a reseller publishing the reports of the brand it sells. No badge; the vendor page notes the earlier concern.
  Each badge shows its reason and the date it was checked.
• Where a report links to the issuing lab's own copy (Freedom Diagnostics, Accumark, Chromate, Vanguard, Kovera, Uzorak, TrustPointe), we open the lab's copy and check it names the compound the vendor claims. Vendors get a "Confirmed with lab" count, or a red "Lab check" badge if a lab's copy names a different compound or the link doesn't resolve. Janoshik blocks automated checks, so its reports are listed as published. Anyone can send us a lab report; it's checked against the lab and reviewed before it appears.

Banned vendors

10 vendors are banned for fake lab reports: reports generated by their website's own code, doctored documents, or reports copied from other brands. Their listings are removed from the shop and the planner, and the reason is shown on the Vendors tab. A ban is lifted only if the evidence turns out to be wrong.

Research summaries and grades

• Sources: 861 papers from PubMed, Europe PMC (including 7 preprints, labelled as not peer-reviewed) and Russian journals via CyberLeninka (22 papers, labelled; summaries are translations). Only papers that actually study the compound are kept: work on a related molecule (full-length thymosin β4 for TB-500, NMN for NAD+), multi-ingredient products, or papers that only measure the body's own levels are excluded or flagged.
• Each paper gets a verdict judged on its own main outcome (positive, negative, mixed, inconclusive), a one-line finding, and where the abstract states them, the sample size, design, key result, the authors' conclusion and limitations. Numbers are checked automatically: a sample size or key number that doesn't appear in the source is rejected.
• Figures are reproduced only from openly licensed full texts (CC0, CC BY, CC BY-SA) via PubMed Central's open-access dataset, with the license and source shown.

Evidence grades

Grades rate how strong the evidence is, not whether the result was good. Every grade comes with a one-line reason on the site.

Only the exact molecule counts. Uncontrolled human reports can't lift a compound above C however many there are, and the whole site is graded with the same rubric in one pass so compounds are comparable.

Clinical trials

Registered trials (4,632) come from the ClinicalTrials.gov API, matched on the compound's name and its code names in each trial's listed interventions. A registered trial isn't evidence of a result; we show status, phase, size, sponsor, countries and whether results have been posted. All trials.

US compounding status

Whether US pharmacies may compound a peptide is read from FDA's own lists (the section 503A categories and the Category 2 safety list) and Federal Register notices, with each status linked to its source and dated. Announcements and advisory-committee votes are shown as pending until FDA actually changes a list or publishes a rule. Status for every peptide.

Doses

We show only amounts given in published studies or official drug labels, each linked to its source and labelled by evidence (approved drug, human trials, animal-only). They aren't recommendations, and animal doses don't translate to people. The calculators do arithmetic for the amounts you enter; they don't suggest any.

Limitations

• Prices and stock can change between collections; always confirm on the vendor's page before buying.
• Finnrick tests are samples. A good score doesn't guarantee every batch, and a vendor's published COAs show what they chose to publish.
• Paper summaries are read from abstracts and, where available, full text. They can miss nuance; follow the links to the papers.
• Research grades and summaries were produced with AI assistance under fixed written rules, then checked automatically against the sources. Errors are possible, and we correct them when found.
• Regulatory status changes; the date checked is shown with each status.