How We Evaluate

Every supplier evaluation on this site is made the same way. Each article states the date its observations were made. This page describes that method.

What we do

• Visit each supplier's live site and read it directly: product pages, certificate pages, shipping and refund policies.
• Open every certificate of analysis that will load, and read it: the lab named on it, the lot number, the method, the date.
• Record prices from the pages themselves, with the product and size, so comparisons are like for like.
• Read community discussion from sources that are not paid to say nice things, and weigh recency.
• Say plainly when something could not be verified, instead of repeating a claim we could not check.

That work is machine work, the same as the articles. An AI system reads a supplier's pages and writes the dossier a record is built from, and before that record enters the ledger the dossier goes through a separate checking pass that works claim by claim, whose verdicts are kept as files with the research rather than published. No record can exist until the language pass that governs every page here passes it. "We" on this page, and the "I" in the dated entries on each supplier record, both name that process. This publication has no editorial staff, and nobody is claiming to have walked a site by hand.

The criteria, in ranking order

1. Certificate practice: is the testing lab named and independently verifiable, is the certificate batch-specific, is the method stated, can a visitor open it.
2. Pricing honesty: visible prices, no decorative strikethrough anchors, no countdown theatrics.
3. Catalog depth and stock reality.
4. Shipping clarity, including storage and transit communication.
5. Community standing.

The Terminus Score printed beside each supplier record is compiled automatically from the dated observations in that record. It counts documented criteria. It is not a review, not a rating, and not an endorsement.

The eight criteria it counts: a certificate of analysis is published; certificates are batch-specific; the testing laboratory is named; that laboratory can be confirmed independently, through an accreditation number or an accreditor's directory; the purity figure names its method, HPLC or mass spectrometry; research-use-only terms are stated; a link to the laboratory's own verification page is offered; and the company behind the store is published, as a legal entity or a physical address. Three recorded gaps subtract from the score: certificates gated behind an account, certificates that could not be read, and a catalog-wide purity claim with no method behind it. Each supplier's record lists all eight and marks which were met on the dates observed.

Correction dated 2026-09-26. The automatic reader that compiles the eight rows above from each record's dated entries was corrected on 2026-09-25 and 2026-09-26, after a supplier reported that two rows on its record contradicted the record's own entries. The correction applies to all 126 records under the eight criteria above, and no dated entry was rewritten to produce it. It moved 82 rows and changed the published score of 57 records. 27 rows moved from not met to met where the reader had missed a laboratory name, research-use-only wording or a link to a laboratory's own verification page that the record states. 55 rows moved from met to not met, 50 of them on the Lab verifiable row, which had been reading met on accreditation wording that the record itself does not confirm through an accreditation number or an accreditor's directory. Scores and board positions moved with the rows. Three Lab verifiable rows are known to still read met on a sentence that denies the evidence; they will be corrected the same way, with a dated note here.

Correction dated 2026-09-26, second note. The three Lab verifiable rows the note above describes as still reading met were corrected the same way on 2026-09-26, together with 33 more rows that were reading met on words their own records deny or on another company's details: 24 Entity published rows, 3 Batch-specific, 3 Method-qualified, 2 Verify link, 1 RUO stated. The correction applies to all 126 records under the eight criteria above, and no dated entry was rewritten to produce it. It moved 36 rows, all from met to not met, and lowered the published score of 29 records, by 1 to 6 each. Rows came off where the words counted were a laboratory's, a registrar's or another company's name or address, a number read as part of an address, an accreditor named inside a denial or only as an example, a method with no purity figure, a certificate belonging to another company, or a denial the reader had read as a statement. Scores and board positions moved with the rows. So far, 27 rows in 19 records are known to still read met on such words; they will be corrected the same way, with a dated note here.

What this method cannot tell you

The method reads what a supplier publishes. It does not measure what is in a vial, and no score here says anything about the contents or quality of a product. A certificate that opens is read, not authenticated: we record the laboratory, lot, method, and date as printed, and only the laboratory that issued it can confirm the document is genuine. How to ask a laboratory that question is in Check a COA in Under Two Minutes. Nothing here measures shipping, order handling, or customer service, because this publication has never placed an order with any supplier. Community discussion is weighed for recency and independence, but we cannot confirm who wrote a post or whether they bought anything. And an observation is true on its date: suppliers change their pages, and a record is only as current as its latest dated entry.

What we never do

Placement on a list is never sold. Suppliers do not see evaluations before publication and cannot pay to change one. When a supplier connected to this publication's owner appears in an evaluation, that connection is disclosed on the page, and the supplier is held to the same criteria as every other entry. A ranking changes when the documentation changes: a supplier that publishes better certificates earns a better evaluation at the next review. That door is open to every supplier equally.