Methodology · v1.1
How we score skincare.
Every product is scored on three pillars: Evidence (does the published research support the mechanism?), Honesty (does the marketing match what's in the jar?), and Value (is the price justified by the formula?).
Each pillar is scored 0-10 using rubric v1.1. Every score has a falsifiability condition — what evidence would change it. Every claim cites a specific source (USPTO, PubMed, FDA, FTC). Every investigation is drafted by Claude (Anthropic's AI) and verified by a human reviewer before publishing.
The three pillars
Evidence
Does peer-reviewed research support the claimed mechanism at the concentrations actually in the product? Brand-funded studies without peer review score lower. PubMed citations score highest.
Honesty
Does the marketing language match the formula and the actual evidence base? Trademarked names presented as patents, vague clinical claims, fear-marketing about competitor ingredients all reduce the Honesty score.
Value
Is the price justified by the formula compared to functional equivalents at lower price points? Score considers concentration disclosure, vehicle quality, and post-patent equivalents where they exist.
Same standards across all brands
The platform commits to applying these standards evenly, regardless of brand prestige, retailer, or how the answer scans:
Top 4 ingredients by INCI order, period. Not the brand-marketed actives. Not the most flattering subset.
Ownership disclosed in the same position on every product. Estée Lauder's La Mer and L'Oréal's CeraVe get the same prominence.
'May not suit' caveats on every product, including high scorers. A 9.1 doesn't exempt a product from contraindication notes.
Falsifiability conditions on every score. The 2.3 and the 9.1 both have to specify what would change them.
No editorial verdicts that read as endorsement. PST observes; it doesn't pick winners.
How investigations are produced
Every investigation is drafted by Claude — Anthropic's AI — using public sources only.
The AI extracts ingredient lists in INCI order, identifies marketing claims, searches USPTO and PubMed for verification, checks the formula against published "may not suit" patterns from dermatology literature, and applies the rubric to produce pillar scores with falsifiability conditions. A human reviewer verifies every citation before publishing.
Each record stores the model version and rubric version used, so any score can be reproduced or challenged.