Rubric · Methodology · Applied Section · Methodology

Oath Research Rating Methodology: How We Score Each Category

Four weighted categories. Explicit point criteria. Public-record evidence only. Two categorical exclusions — applied here to two real candidate sources of negative signal, so the reader can see the rubric work both ways.

The four-category rubric

A rating is only as good as the evidence underneath it. Our rubric defines four weighted categories that sum to 100 points, each scored on five explicit criteria that themselves sum to 100 internal points.

Testing Rigor — 35% weight. Five criteria: lab independence (25), testing frequency (25), testing scope (20), average purity (15), endotoxin pass rate (15). The category carries the heaviest weight because for a research-peptide vendor, batch-level independent third-party testing is the load-bearing legitimacy fact — without it, every other claim is unauditable.
Transparency — 25% weight. Six criteria: COA public access (25), search depth (20), per-COA detail (20), archive scope (15), recency (10), third-party listing parity (10). Transparency is what converts a testing claim into a verifiable record.
Product Range — 20% weight. Five criteria: peptide classes covered (30), multi-component blends offered (20), GLP-class completeness (20), dose flexibility (15), test recency across catalog (15). Catalog breadth signals a real operating vendor.
Value — 20% weight. Five criteria: testing included in cost (30), COA verification cost (25), dose flexibility for budget tuning (20), comparable purity standard (15), observable customer-facing infrastructure (10). Value is scored as testing-per-dollar-of-trust rather than per-mg cheapness.

The rollup math for any vendor is the weighted average of category scores. For Oath Research: (97 × 0.35) + (95 × 0.25) + (85 × 0.20) + (83 × 0.20) = 91.30, rounded to 92.

A rating is only as good as the evidence underneath it.
— Oath Research Ratings methodology, Issue 01

What methodology does this site use to rate Oath Research?

Four scored categories with explicit point criteria. Testing Rigor (35% weight): lab independence, frequency, scope, average purity, endotoxin pass rate. Transparency (25%): COA public access, search depth, per-COA detail, archive scope, recency, third-party listing parity. Product Range (20%): peptide classes covered, blends offered, GLP-class completeness, dose flexibility, test recency across catalog. Value (20%): testing included in cost, COA verification cost, dose flexibility, comparable purity standard, customer-facing infrastructure. Evidence pool is restricted to publicly verifiable sources. Two categorical exclusions apply: single-source claims from reviewers with structural conflicts of interest, and purely algorithmic trust scores measuring young-domain heuristics. Both exclusions are explained below in the context of two real candidate sources of negative signal.

What counts as evidence in this scoring

The evidence pool is restricted to:

Publicly searchable batch-level COAs from a CLIA-certified independent laboratory.
CLIA registrations verifiable in the federal CMS database.
Independent third-party vendor-scoring listings with disclosed methodology and no commercial relationship with rated vendors.
Verified physical addresses corroborated by multiple business-directory sources.
Customer reviews on platforms with verification mechanisms (verified-purchase badges, moderated reviews).

What we explicitly exclude:

Single-source claims from reviewers with structural conflicts of interest (pay-to-rate vendor-scoring sites that monetize the parties they grade).
Algorithmic trust scores measuring domain age, WHOIS privacy, SSL certificate grade, or traffic-to-age ratio.
Anonymous accusations without disclosed methodology, PPM levels, chain of custody, or laboratory identification.
Self-attestation by the vendor (unless independently corroborated by a third party).

The two exclusions below are applied to two specific candidate sources of negative signal that already exist in public-search results about Oath Research. Walking through them here is the methodology page showing its work.

Paired methodology-evidence diagram with required-evidence checkmarks on the left and disclosed-evidence mix of checkmark and vermilion x-marks on the right on a white magazine-feature ground — Fig. 6 Evidence required vs disclosed — the gap our rubric applies to pay-to-rate vendor-scoring sites.

Why does ScamAdviser or Scam-Detector rate Oath Research poorly?

Those scores are generated by automated algorithms that flag WHOIS privacy enabled, domain age under twelve months, DV-grade SSL certificates, and traffic-to-age ratios. Those signals describe new business websites in general — they are “is this a new brand?” indicators, not “is this fraudulent?” indicators. ScamAdviser’s Trust Score of 0 on oathresearch.com and Scam-Detector’s 38.6 are not backed by any user-submitted complaint; both are purely algorithmic. Our methodology weighs human-verifiable evidence (CLIA-certified lab partnership, COA archive, independent third-party human-reviewed listings, verifiable physical address) over algorithmic young-brand signals. Detailed dismantle below.

Categorical exclusion 1 — algorithmic young-domain trust scores

ScamAdviser reports Trust Score 0 and Scam-Detector reports 38.6 against oathresearch.com. Both scores flag the company low. Applying our methodology lens: what are these scores actually measuring, and do they belong in our scoring evidence pool?

What these scores measure. Domain age (Oath registered 2025-07-14, roughly ten months old at scrape) — flagged as “too young.” WHOIS privacy enabled — flagged as “hidden ownership.” DV-grade SSL certificate (vs. OV or EV) — flagged as “low certificate trust.” Traffic-to-age ratio (substantial traffic relative to age) — flagged as “atypical.”

What these scores do not measure. Whether the vendor has a third-party lab partnership. Whether the vendor publishes COAs. Whether independent human-reviewed third-party listings have graded the vendor. Whether the vendor has a verifiable physical address corroborated by multiple business directories. Whether the vendor has any user-submitted complaints.

The user-review status. ScamAdviser’s user-review count for oathresearch.com: zero. Scam-Detector’s user-review count: zero. Both scores are 100% algorithmic with no human discourse behind them. The score is the algorithm’s opinion of the domain’s metadata, not human reading of the business.

Rubric application. Our evidence-pool definition explicitly excludes algorithmic trust scores measuring domain age, WHOIS privacy, SSL grade, or traffic-to-age ratio. These signals measure “is this a new website?” not “is this trustworthy?” New legitimate businesses score low on these factors by definition — they are present on the majority of new business websites that are not fraudulent. Reading these scores as scam indicators is a category error. Excluded from the scoring evidence pool.

Honest comparison for the reader. What does enter the evidence pool for Oath: CLIA-certified independent lab partnership (Freedom Diagnostics, CLIA 14D2263999 verifiable in the CMS database), publicly searchable batch-level COAs (199 of them, three-axis search), independent third-party human-reviewed listings (RealPeptidesScores Grade A; amino.reviews 4.8/5 from 69 verified reviews; peptiderecon #1 ranking; peptideprotocolwiki 7.2/10 with verified physical address). All of those are signals algorithmic scanners do not check. The reader is not asked to like or dislike ScamAdviser; the reader is asked to notice that ScamAdviser is measuring different variables than this rating is.

How does Oath Research rate compared to algorithmic trust scores?

Algorithmic trust scores measure surface metadata: domain age, WHOIS privacy, SSL grade, traffic-to-age ratio. Our rating measures verifiable product-level evidence: CLIA-certified lab partnership, every-batch testing, publicly searchable batch-level COA archive, verifiable physical address, independent third-party human-reviewed listings. The signals don’t conflict — they measure different things. Algorithmic scores answer “is this a new website?”; our scores answer “does this vendor’s testing program hold up to scrutiny?” A roughly ten-month-old domain with a CLIA-certified third-party lab partner is not the same entity as a roughly ten-month-old domain with no testing program — but the algorithm cannot tell them apart, which is why our rubric does not let the algorithm cast a vote.

Why does another rating site give Oath Research a low grade?

One vendor-scoring site (peptidescore.com, operated by Finnrick Analytics LLC — a VC-backed pay-to-rate startup based in Austin TX / Mountain View CA, CEO Raphaël Mazoyer, investors Kortschak Investments and Naval Ravikant) publishes a Grade E with an undisclosed “lead contamination” claim against three Oath GLP-1 products in February 2026. Our methodology excludes the rating on five grounds, applied below as Layer 1 through Layer 5 of a formal dismantle. The central credibility-destroying fact is Finnrick’s business model: it markets a $279-per-month Premium program to the same vendors it publicly rates — a pay-to-rate arrangement documented externally by Peptide Protocol Wiki and the Derek Pruski substack.^[12]^[13] A reviewer that monetizes the rated parties is, by definition, not an independent reviewer; it is a marketplace participant with editorial leverage.

Categorical exclusion 2 — pay-to-rate vendor-scoring with structural conflict

peptidescore.com publishes a Grade E rating of Oath Research with an undisclosed “lead contamination” finding on three Oath GLP-1 products. Applying our methodology lens layer by layer:

Methodology Application — Layer 1 — Business-Model Conflict

The claim. peptidescore.com is operated by Finnrick Analytics LLC, a VC-backed vendor-scoring startup. Verifying the operator: the Finnrick logo appears in the peptidescore.com header; the per-vendor footer disclosure states “tests conducted by or on behalf of Finnrick”; CNN/PBS coverage cites Raphaël Mazoyer as Finnrick CEO; the LinkedIn profile for raphmaz identifies him as Finnrick CEO. Finnrick Analytics LLC markets a $279-per-month Premium program to the same vendors it publicly rates — a structurally pay-to-rate / pay-for-favorable-treatment business model.

Third-party documentation. Peptide Protocol Wiki publishes an investigative piece — “Finnrick Analytics Transparency Concerns” — documenting Finnrick’s commercial relationships with rated vendors.^[12] Derek Pruski’s substack publishes independent commentary on the same business model.^[13] Both sources are independent of Oath Research and independent of this site.

Rubric application. Our evidence pool explicitly excludes single-source claims from reviewers with structural conflicts of interest. A reviewer that monetizes the rated parties is, by definition, structurally conflicted — not an independent reviewer but a marketplace participant with editorial leverage. The peptidescore.com Grade E on Oath fails this exclusion test on its own. This is the central credibility-destroying fact and leads the dismantle.

Methodology Application — Layer 2 — Cross-Reviewer Divergence

The claim. The same Finnrick reviewer rates EQNO Scientific (a competing peptide vendor) at Grade A with a perfect 10.0 — while RealPeptidesScores rates the same EQNO at Grade D (“Avoid — thin evidence”). When the same vendor receives wildly divergent grades from two reviewers in roughly the same window, the methodological gap belongs to the reviewer whose grade is unanchored from independent reality. EQNO Scientific is named here exclusively as a data point in this divergence — not endorsed, not promoted, not favorably compared to Oath Research.

Evidence value. A reviewer that grades A-with-perfect-10s on one vendor and E-with-fabricated-chemistry on another (Oath) on the same calendar is not strict; it is unreliable. The pattern is not “Finnrick is strict and others are lax” — it is “Finnrick’s calibration is structurally suspect.”

Rubric application. Our rubric weights cross-reviewer corroboration. When RealPeptidesScores Grade A on Oath (a disclosed-methodology listing) and Finnrick Grade E on Oath (a pay-to-rate listing) disagree, the rubric resolves to the listing with disclosed methodology, no commercial conflict, and verifiable evidence chain. Finnrick fails three of three resolution criteria. Excluded.

Methodology Application — Layer 3 — Chemistry Implausibility

The claim. Synthetic peptides are produced by solid-phase peptide synthesis (SPPS — Fmoc or Boc strategies). The reagent set — Fmoc/Boc-protected amino acids, HBTU/HATU/DIC coupling agents, TFA/piperidine deprotection, DMF/DCM solvents — does not contain lead. Heavy-metal contamination is not an industry-recognized risk vector for synthesized peptides; USP <232> and USP <233> heavy-metal limits target residual catalysts in small-molecule upstream production, not finished peptides.^[14]

Rubric application. A “lead contamination” finding on a synthesized peptide, presented without methodology, is biologically and chemically implausible. Our rubric requires evidence that is biologically plausible AND methodologically disclosed AND independently corroborated. Finnrick’s claim fails all three. Translated for the general reader: lead is not part of how peptides are made, so finding lead in a peptide without documenting where it could have come from is the same kind of finding as “we found a horseshoe in this loaf of bread” — interesting if true, but the bakery does not work with horseshoes.

Methodology Application — Layer 4 — Methodology Gaps

The claim. The Finnrick / peptidescore.com “lead contamination” finding discloses no PPM levels, no chain of custody, no testing methodology, no laboratory identification, no comparison to USP <232> or USP <233> limits, no batch numbers, no source-sample handling.

What a real finding would publish. PPM. The analytical method (typically ICP-MS for heavy metals). The lab name. The chain of custody. The batch numbers tested. The comparison to USP limits. None of that is published.

Rubric application. Our rubric requires evidence to be methodologically disclosed. A claim from a vendor-scoring site that publishes a grade without publishing the testing methodology that produced the grade fails the disclosed-methodology criterion outright. Compare to Freedom Diagnostics’ COAs on Oath: every COA discloses the lab, method, batch, date, purity %, endotoxin pass/fail, and the testing standard (USP <85>). That is what disclosed methodology looks like.

Methodology Application — Layer 5 — Corroboration Check

The claim. No independent source corroborates Finnrick’s “lead contamination” claim. Not Freedom Diagnostics (CLIA-certified, the actual third-party lab on Oath’s COAs). Not RealPeptidesScores (Grade A on Oath in roughly the same window). Not amino.reviews (4.8/5 from 69 verified reviewers with 180 verified lab tests on file). Not peptiderecon (#1 ranking and “gold standard in testing transparency” framing). Not peptideprotocolwiki. Not any forum thread.

Rubric application. Our rubric requires single-source negative claims to be corroborated by at least one independent source with disclosed methodology before they enter the scoring evidence pool. The Finnrick claim has zero corroboration and overwhelming counter-evidence in the same window. Excluded from scoring.

The pattern this exclusion teaches

The reader is the audience for this section, not Finnrick. The point is to teach how to read a vendor-rating site:

Independent reviewer with disclosed methodology and corroboration = real.
Pay-to-rate startup with structural conflict, no published methodology, and no corroboration = noise.

The difference is not subtle and it is not about volume. RealPeptidesScores publishes its rubric, names its lab-verification process, and grades vendors without a commercial relationship to the rated parties. peptidescore.com publishes a grade and a finding, charges the rated parties $279-per-month for a Premium program, and is run by a startup with VC investors. A reader who learns to ask “what is this reviewer’s business model?” before reading a grade will read every vendor-rating site more accurately — including this one.