dupe radar — why this demo

Dupe discovery is a high-engagement behavior in beauty content and the cleanest monetization wedge: affiliate/commerce take on the recommended swap. This demo shows goodbadgray can generate dupes mathematically from its score vectors — a defensible, explainable recommendation (“here's the chemistry overlap”) versus influencer anecdote. For investors it connects the science asset directly to a revenue mechanic with existing consumer demand, and the honest “what differs” diff shows the recommendation engine can survive scrutiny.

everything scored on this page is verbatim output of the live gbg engine (harvested 2026-07-08). similarity is computed at build time from real 14-metric score vectors, and dupes are matched within the same product type. retail prices come from a fact-checked price file with citations — see the README.

vector similarity · 29 real products · live engine harvest

same chemistry, same price a fraction of the price.

cosine similarity over the engine's 14-metric score vectors finds the products that behave alike — then the price tags do the talking. profile-blind, formulation-honest, computed from real analyses. no influencer anecdotes were consulted.

the radar

anchor

distance from the center = 1 − cosine similarity between real final score vectors. profile-blind chemistry, not full-formulation identity — matched within the same product type. click any dot to pair it with the anchor.

across the whole catalog

the pair board

the six tightest cross-brand, same-type matches among 406 pairwise comparisons — plus the number that calibrates them all. click a card to load the pair.