A rank is built only from a full dossier, by this registry's own rule

ON THIS PAGE

1. Key findings
2. Four tiers, read directly
3. What the rule actually forbids
4. The measured gap between screened and known
5. What sitting in a lower tier does not mean
6. What a reader can ask of any other ranking site
7. Limitations and open questions
8. Sources

A supplier can sit in this registry for months without ever appearing in a published rank, and that gap is not an oversight. It is the point. Observation and judgment are different tiers here, and the second is never built from the first alone.

Four tiers, read directly

This registry's own documentation, read on 2026-09-19, names four tiers in order. Tier 0, discovered, holds one line per domain, a name if one is known, and the source that named it; holding an entry at this tier asserts nothing except that a source named the domain. Tier 1, screened, holds machine-checkable signals gathered by an automated pass: whether the domain answers, what its homepage's own text says, whether it carries research-use-only language or a certificate-of-analysis mention. Tier 2, evaluated, is a full evidence dossier, the kind of document this publication's own ledger holds for a supplier once someone has actually read its pages closely rather than run a fixed automated check against them. Tier 3, ranked, holds published scoresheets, and the documentation states plainly that a Tier 3 rank is built only from Tier 2 dossiers. Nothing at Tier 0 or Tier 1 alone is sufficient to place a supplier in a rank.

What the rule actually forbids

This publication's own review rubric, the document a fresh adversarial reviewer reads before checking any draft, states the same rule as a condition that fails a piece outright rather than as a guideline: "A ranking built on anything other than the ledger (a forum, a vendor's own claim, a general impression) is blocking on its face." The same section states that a stated rank, score, or price must match the ledger exactly, and a number the ledger no longer supports is either corrected from the ledger or the piece does not run. The rule is not only about accuracy. It is about which tier of evidence a rank is allowed to draw from at all, and a rank built from anything short of a Tier 2 dossier fails this test regardless of whether the number itself happens to be right.

The measured gap between screened and known

The distance between tiers is not just asserted in this registry's documentation. It is measured. Before 2026-08-23, Tier 1 screening walked its domain list in plain file order, and after 3,800 domains read that way, only 18 carried any peptide signal at all, and none carried research-use-only language. After the screening order changed to read peptide-specific sources first, the documentation records that the first 300 domains under the new order returned 225 alive, 107 carrying a peptide signal, 56 carrying research-use-only language, and 80 carrying certificate-of-analysis language, a yield the documentation itself states as roughly 46 times higher per domain screened than the earlier pass. Both passes are still Tier 1. Neither one produced a dossier, and neither one produced a rank; what changed was how quickly Tier 1 found the domains worth a Tier 2 look, not what any of those domains were found to be doing.

What sitting in a lower tier does not mean

A domain that has only reached Tier 0 or Tier 1 in this registry is not thereby judged, favorably or otherwise. Tier 0 states only that a source named it. Tier 1 states only what a fixed automated check found on one homepage on one date. Neither tier says anything about whether the business behind that domain is operating honestly, poorly documented, or anything else a Tier 2 dossier might eventually establish. What a lower tier actually withholds is coverage, not a verdict: a supplier this registry has not yet screened, or has screened but not evaluated in depth, is invisible to any rank it might otherwise have earned a place in, and invisible is a different fact from clean. A supplier absent from every published rank here could be a supplier nobody has looked at closely yet, not a supplier that was looked at and cleared.

What a reader can ask of any other ranking site

The test this registry applies to its own ranks is one a reader can ask of any other site that ranks suppliers in this category. What tier of evidence does a listed rank actually draw from, a full evaluation, a one-time automated check, or nothing checkable at all. Does the site say so anywhere, or does a rank appear with no stated evidentiary floor beneath it. And does the site's own rule, if it states one, actually bind what gets published, or is it a stated ideal a specific rank quietly falls short of. A rank that cannot answer where its numbers came from is functionally the same as a rank built on a forum thread or a vendor's own claim, whatever tier structure the site claims to run underneath it.

Limitations and open questions

This piece describes this registry's own tier structure and rubric as read directly from its own documentation on 2026-09-19; both can be revised after that date, and a reader checking this description later should verify it against the current source rather than this piece. The 46-times yield figure and the tier definitions are read directly from this registry's own stated numbers rather than independently recomputed for this piece, and this piece takes no position on how many suppliers currently sit at each tier or on whether every published rank on this site in fact draws only from Tier 2 evidence, which is a separate audit this piece does not perform.

A rank is the last thing this registry is willing to say about a supplier, not the first, and getting there requires clearing three tiers of evidence a bare listing does not.

Sources

This publication's own registry documentation, registry/README.md, and its own review rubric, REVIEW-PROTOCOL.md, both read directly on 2026-09-19, for the tier definitions, the screening-order yield figures, and the ranking rule quoted above.