Vendor Ratings

Vendor directory

Scores are shown only when a vendor has passed the minimum evidence rule described below. Until then the vendor is listed as Not Yet Rated, which means there is not enough verified public evidence to publish a fair score yet. It does not mean the vendor scored zero.

Review a seller

Bought from one of these vendors? Your review is the one source that can cover every vendor, including the ones no public site talks about. Reviews are moderated before they count, are limited to one per person per seller, and count as their own source category. A review never publishes a score by itself: the seller still needs 15 mentions across 2 or more source categories before any score appears.

How these ratings are calculated

This page is the full method, published so any vendor or shopper can check exactly how a score was produced. The same rules apply to every vendor, and the same system computes every score.

1. Where the evidence comes from

The system collects public material that names a vendor from the sources below. A mention only counts toward a score when it can be attributed to a source and a date, and when it is about the seller and its service. Product listings, advertisements, and the vendor's own marketing copy are not evidence and are never scored.

• Reddit: Public community discussions in which a vendor is named and its service is discussed.
• Finnrick Analytics: Independent laboratory testing data: published analytical test results that name a vendor and its products.
• DisclosedLabs: Independent laboratory testing evidence: published certificates of analysis and a vendor trust score, collected through the free DisclosedLabs API.
• PeptideCritic: Published vendor reviews from an independent review publication covering peptide vendors and their service.
• On-site seller reviews: Reviews submitted through the form on this page. Each one is moderated before it counts, is limited to one per person per seller, and counts as its own source category.

Independent testing is one evidence source above. Shoppers who want a test of their own can use the peptide lab testing directory, which lists third party laboratories with published prices and turnaround times.

2. Automated language analysis

Each collected mention is processed by an automated language analysis system, not scored by hand. The system classifies the sentiment of the text as positive, neutral, negative, or mixed, and tags what the mention is about, including shipping, support, refunds, accuracy and quality, packaging, payment, fraud, satisfaction. Those tags are what allow shipping problems, support problems, and refund problems to affect different parts of the score instead of being blended into one opinion.

Mentions that contain medical claims are flagged for compliance during analysis. Flagged content is excluded from publication, and no review text is republished on this site.

3. The weighted scoring model

Six components are combined into one score on a 0 to 5 scale. Where a component has no qualifying evidence for a vendor, its weight is redistributed across the components that do, so a missing component never silently counts as zero.

4. Ratings from any scale, normalized to 0 to 5

Sources do not all rate the same way. A numeric rating on any scale, for example 5 stars, a 10 point scale, or a percentage, is converted proportionally onto the common 0 to 5 scale before it enters the model. A 4 on a 5 point scale and an 80 on a 100 point scale therefore count the same.

5. Recent evidence counts more: the 180 day half-life

A vendor's service can improve or decline, so older evidence fades. Each mention is weighted by recency with a half-life of 180 days. A mention published today counts at full weight, a mention from 180 days ago counts at half weight, and a mention from a year ago counts at roughly a quarter weight. Old praise cannot carry a vendor forever, and an old problem stops dominating once newer evidence replaces it.

6. Protection against small samples

A vendor with two glowing reviews should not outrank a vendor with two hundred good ones. Scores are therefore smoothed with a Bayesian method that pulls every score toward a neutral midpoint. The pull is strong when evidence is thin and fades as a vendor's own evidence grows, so a published score reflects a vendor's own record, in proportion to how much verified record there is.

7. Duplicate and fake review protections

Every mention is fingerprinted before scoring. Mentions with the same fingerprint, including the same text posted in more than one place, are deduplicated and counted once. Patterns that suggest manipulation, such as sudden bursts of similar reviews, raise a manipulation flag. While a flag is open, publication of that vendor's score is frozen until the evidence is cleared, so a score is never published on the strength of suspect reviews.

8. The minimum evidence rule

A score is published only when a vendor has at least 15 qualifying mentions drawn from at least 2 different source categories. Below that line the vendor shows as Not Yet Rated. This gate exists so that no vendor is judged, positively or negatively, on a handful of comments.

9. Weekly rescans, and scores that move both ways

The full collection and scoring run repeats on a weekly cycle for every vendor. When the new run finishes, the published score is replaced with the recomputed score and the last scanned date on this page updates. Scores can move up or down. A vendor that fixes its shipping earns a better score as new evidence arrives, and a vendor whose service slips will see its score fall.

Vendor ratings are provided for shopping information only. They are computed from public third party material by an automated system, and they do not represent product efficacy, safety, purity, or regulatory approval. All products on this marketplace are sold for research use only.