Vendor Trust Score: how we compute it.
The Pepticker Trust Score is a composite 0–100 number computed mechanically from the data we already publish about every vendor. Methodology is public, weights are listed, the algorithm lives in source control. We do not sell. We disclose every affiliate relationship inline with the rankings they inform. The score is a decision aid, not an endorsement.
What goes into the score
Five components, weighted to total 100. Each component is computed purely from a field we already publish on the vendor directory. No private data, no opaque ratings, no editorial vibes.
- 01
Testing transparency
40 ptsSource:testingScore + testingMethods + testingLabPublic 7-category testing rubric grade plus disclosed testing methods plus a named third-party lab. The full range is gated on the vendor publicly naming the tests they run; a high rubric grade without disclosure caps at 18 of 40.
Worked example: a vendor with
testingScore = 6.0, four disclosed testing methods (HPLC purity, mass-spec identity, endotoxin, sterility), and a named lab earns round(6.0 · 30/7) + 4 + 6 = 26 + 4 + 6 = 36 of 40. A vendor withtestingScore = 6.0buttestingMethods = []caps at round(6.0 · 18/7) = 15 of 40 — non-disclosure costs you 21 points. - 02
Operational & regulatory standing
25 ptsSource:riskStatus + auditNotes + latestCoaDateReflects FDA actions, payment-processor issues, ownership uncertainty, closures, and editorial audit notes the vendor-risk monitor has surfaced. A recent published COA earns a small bonus on top of the operational base.
Worked example: an operational vendor with no audit notes and a recent COA earns the full 22 + 3 = 25. A vendor with
riskStatus = "watch"and one audit note about an unverified second storefront earns 6 of 25. - 03
Operational stability
15 ptsSource:foundedYearVendor age. 5+ years of operation earns full points; 1 year or less earns minimal credit. Absent foundedYear earns a neutral default; absence of data is not proof of risk.
Worked example: a vendor with
foundedYear = 2019(7 years old as of 2026) earns the full 15 of 15. A 2025-founded vendor earns 8. - 04
Verification tier
10 ptsSource:tier (1/2/3)Editorial tier assignment based on documentation completeness, response to corrections, and presence of verifiable business address. Tier 1 vendors meet the highest bar.
Worked example: a tier-1 vendor earns 10 of 10; tier-2 earns 6; tier-3 earns 2.
- 05
Price discipline
10 ptsSource:discountCode + discountPctHeuristic check on aggressive discounting. Reasonable codes (<=25%) earn full points; 25-40% earns half; >40% off list often signals inventory clearance or churn and earns zero.
Worked example: a vendor running a 15% off code earns the full 10 of 10. A 35% off code earns 5. A 50% off code earns 0 — heavy discounting often signals inventory clearance or vendor churn.
On the regulatory component: Pepticker does not currently maintain a dedicated FDA-warning-letter field on the Vendor record. Our riskStatus field captures FDA actions, payment-processor issues, ownership uncertainty, and operational closures together. If a vendor receives a public FDA warning letter we mirror it as riskStatus = "watch" or "degraded" and surface it in audit notes.
From score to letter grade
Grade boundaries are tight on purpose. Most vendors land in B/C, not A. That is a feature, not a bug — a methodology that handed out A+ to half the directory would not be a useful decision aid.
| Grade | Score | Meaning |
|---|---|---|
| A+ | 90-100 | Highest confidence. Full testing rubric, named lab, operational, established history. |
| A | 80-89 | High trust. One or two minor signals short of A+ — usually a missing testing-method line item or a very recent founding date. |
| B | 70-79 | Moderate trust. Operational and broadly compliant, but the testing-transparency or audit-note signals are incomplete. |
| C | 60-69 | Median band. The vendor either lacks public testing methods or carries audit notes that affect the regulatory-standing component. |
| D | 50-59 | Caution. Multiple weak signals — likely a 'watch' status combined with thin testing disclosure. |
| F | 0-49 | Avoid until further evidence. Often closed, degraded, or carrying multiple audit notes. |
A decision aid, not an endorsement
The Trust Score is a comparison aid. It is not a recommendation, an endorsement, or a substitute for your own due diligence. Higher trust does not automatically mean better price — consider both. A B-graded vendor with the lowest $/mg may be the right pick for a familiar compound; an A-graded vendor with a 25% premium may be the right pick for an unfamiliar one. The score reframes the question, it does not answer it.
We do not believe a single number can fully describe a vendor. We publish the Trust Score because the alternative — an unranked directory — rewards aggressive marketing over quiet competence. The directory remains sortable by live sales, testing rubric grade, and country.
When the score moves
The score is recomputed on every site build. Vendor data is reviewed on a rolling cadence; price data refreshes every 4–6 hours and feeds the broader risk monitor. When a vendor publishes new testing methods, names a lab, or updates their COA, the testing-transparency component moves up. When the vendor-risk monitor sets riskStatus to watch or above, the regulatory component moves down. We do not freeze scores — the directory is a live snapshot.
Where the model is imperfect
- Operational stability favors older vendors. A well-run vendor founded in 2024 cannot earn the full 15 stability points until 2029. This is intentional — vendor history tracks survivorship signal — but it does penalize new entrants with otherwise-strong testing programs.
- Price discipline is a heuristic. Aggressive discounting often correlates with vendor churn or inventory clearance, but not always. Loss-leader pricing on a single SKU does not necessarily indicate a stressed business. We weight this only 10 points for that reason.
- Tier assignments are editorial. The 1/2/3 tier is assigned by the editorial team based on documentation completeness and verifiable business presence. It is the most subjective input in the model, which is why we cap its weight at 10.
- The score does not measure customer service. Shipping speed, support responsiveness, and packaging quality are real trust signals we don't currently quantify. They are also the signals most easily gamed via paid reviews, which is why we have avoided importing them.
- FDA actions are surfaced indirectly. We track FDA warning letters via
riskStatusand audit notes rather than a dedicated field. A future revision may break this out so the regulatory component can carry distinct weights for FDA actions vs. payment-processor issues.
Questions or corrections? corrections@pepticker.com. The algorithm source lives in lib/trustScore.ts in the public repository.
Related reading: Pepticker methodology · Every vendor we track · About Pepticker.