Version 1.0
How vendors are scored
Every score badge on this site links here. A score whose derivation you cannot check in one click is an assertion, not a rating — the same rule the evidence rubric already enforces for claims.
Last updated
The eight dimensions
Each vendor score decomposes into these, each on a 0–10 scale. 129 of 129 vendors carry the full decomposition; the rest state that the decomposition is not on file rather than pretending one exists.
COA quality
Whether certificates of analysis exist, who produced them, and whether a reader can check them without asking.
What raises it
Third-party certificates, published openly, batch-specific, verifiable at the issuing lab.
What lowers it
In-house-only paperwork, certificates supplied on request, or none at all.
It cannot say the vial you receive matches the paper — a COA describes the batch that was tested, nothing more.
Transparency
How much the vendor states plainly and documents: testing regime, named labs, policies, provenance.
What raises it
A named testing lab, published results, stated shipping and refund terms.
What lowers it
Claims that cannot be traced to a document.
It cannot verify a claim beyond the documents offered for it.
Community reputation
What buyers report across public communities over time.
What raises it
Consistent independent reports of orders arriving as described.
What lowers it
Patterns of non-delivery, failed third-party tests, or silence when challenged.
It is aggregated hearsay, not a measurement — the softest input on this list, and read that way.
Product range
Breadth and availability of the catalogue.
What raises it
A wide, in-stock range.
What lowers it
A thin or chronically out-of-stock list.
Breadth says nothing about the quality of any single product.
Pricing
Price level against the market for comparable products.
What raises it
Low prices that do not coincide with missing testing.
What lowers it
Prices well above market without a stated reason.
It cannot capture your total cost — shipping, fees and exchange rates move per order.
Shipping speed
Dispatch and delivery times as stated by the vendor and reported by buyers.
What raises it
Fast, tracked dispatch to the vendor's stated destinations.
What lowers it
Slow or untracked dispatch, or stated times buyers contradict.
It cannot promise your parcel's fate — customs and carriers are outside any vendor's control.
Website experience
Whether the storefront states what a buyer needs — contents, testing, terms — and works.
What raises it
Clear product pages, accessible documents, a working checkout.
What lowers it
Broken pages, missing information, dark patterns.
A polished storefront is not a quality signal for what is in the vial.
Payment flexibility
How many reasonable ways there are to pay.
What raises it
Cards plus alternatives.
What lowers it
A single payment rail, or rails that remove all recourse.
It says nothing about the safety of the goods.
How the overall score composes
The overall score was carried from the source directory, and it is not the arithmetic mean of the eight dimensions — we checked at merge time rather than assuming. The source applied a weighting it did not publish. Until the composition is re-derived and published here, treat the eight dimensions as the auditable unit and the overall as a carried summary. That is an honest limitation of ported data, stated instead of smoothed over.
Paid placement
Sponsored slots exist on this site, under these rules, all of them checkable:
- A sponsored slot appears only inside a strip labelled Sponsored, with the disclosure inline at the placement — never only on a policy page.
- Sponsored outbound links carry
rel="sponsored". - Scores are not for sale. Payment never changes a score, and a score-ordered list is never reordered, seeded or trimmed for a sponsor. The comparison table contains no paid placement at all.
- Every placement has an end date on file, and the build drops a lapsed placement at the first rebuild after it expires.
- The build fails mechanically if a sponsored strip renders without its inline disclosure.
An earlier version of our own build standard ruled out paid placement entirely. We now sell clearly marked slots instead, under the rules above; the reasoning is recorded in the project's decision log, and this page is the binding public statement of it.
Coupons
A coupon is published only with the code, the discount, a source, and the date we saw it. Expired codes are removed at build. A vendor with no code on file says so — a blank would invite you to assume one exists. 0 codes are on file today, across 129 vendors.
Separately, an offer a storefront displays on its own homepage — a welcome discount, a newsletter signup offer — may be quoted on the vendor's record as a dated observation. It is reproduced as displayed, never validated, and never counted as a coupon code. Sale prices in a product carousel are the store's own regular-versus-current figures from its public feed, shown with the date we read them.
Where the data comes from
These records were built for an earlier directory the operator ran, and carried over here: 129 vendors and 11 labs, captured 2026-08-12. Descriptions and pros/cons are reproduced verbatim (markup and emoji stripped) and are labelled as carried copy wherever they appear. Every record carries two dates — when the source last verified it, and when we captured it.
Nothing on this host fetches live data; this site's security policy forbids the browser from talking to any other host. Freshness arrives only when the dataset is re-fetched, re-merged and the site rebuilt — which is why every price and shipping fact renders with its observation date rather than an implied "now".
Verification cadence and corrections
Records are re-verified on a rolling basis; the per-record dates are the honest statement of where that stands. If a fact here is wrong or stale, the corrections path applies to shop records exactly as it does to compound records: name the field, name the source, and it gets fixed or the absence gets declared.