30,553
Total ingredients
Compounds in our structured safety database
SELECT COUNT(*) FROM ingredients
3.1× ChemComply (10,000 — the only competitor who publishes a count)
Live from benda-ingredients D1
38,411
Unique NOAEL study records
Deduplicated toxicological study records with No-Observed-Adverse-Effect-Level values from REACH, SCCS, COSMOS, FDA, CIR, and Health Canada
SELECT COUNT(*) FROM noael_studies
Coptis Tox has ~3,000 flat NOAEL values. We have 12.8× more unique study records.
Live from benda-ingredients D1
3,608
Unique CAS numbers with NOAEL data
Distinct substances (by CAS registry number) with at least one study-level NOAEL
SELECT COUNT(DISTINCT cas_number) FROM noael_studies WHERE cas_number IS NOT NULL
Live from benda-ingredients D1
57,580
Confidence-scored MoS calculations
Pre-calculated Margin of Safety values across 20 SCCS product categories — each tagged with a data confidence tier (SCCS-validated, REACH-sourced, or LOAEL-corrected). Quarantined data excluded from this count.
SELECT COUNT(*) FROM calculated_mos WHERE data_confidence NOT LIKE 'quarantined%' AND data_confidence NOT LIKE 'suspect%'
No competitor has pre-calculated MoS at this scale — at any price.
Live from benda-ingredients D1
3,080
SCCS-validated MoS calculations
MoS values where our NOAEL was verified against the actual SCCS published opinion — the gold standard for cosmetic safety assessment.
SELECT COUNT(*) FROM calculated_mos WHERE data_confidence IN ('sccs_validated', 'trusted_source')
Live from benda-ingredients D1
234
Gold-tier substances
Substances with measured dermal absorption + full SCCS methodology applied
SELECT COUNT(DISTINCT cas_number) FROM calculated_mos WHERE da_tier = 'gold'
Live from benda-ingredients D1
3,193
Silver-tier substances
Substances with NOAEL data + SCCS default 50% DA
SELECT COUNT(DISTINCT cas_number) FROM calculated_mos WHERE da_tier = 'silver'
Live from benda-ingredients D1
230
Substances with measured dermal absorption
Distinct CAS numbers with at least one empirical DA measurement
SELECT COUNT(DISTINCT cas_number) FROM dermal_absorption
HuskinDB public database: 73 substances. We have 3.2× more.
Live from benda-ingredients D1
860
Dermal absorption records
Individual DA measurements across 5 cross-referenced databases
SELECT COUNT(*) FROM dermal_absorption
Live from benda-ingredients D1
354
SCCS/SCCP opinion PDFs fully extracted
European Commission scientific opinions on cosmetic ingredients, parsed and structured
SELECT COUNT(DISTINCT source_pdf) FROM sccs_pdf_extractions
Live from benda-ingredients D1
7,370
Quantitative records from SCCS opinions
Individual extracted data points (NOAEL, DA, MoS, SED, conclusions) from the PDFs
SELECT COUNT(*) FROM sccs_pdf_extractions
Live from benda-ingredients D1
700,764
Rows traceable AND clickable to a primary source on disk
KAIROS #13 binary rule (April 14, 2026): every row across 14 fact tables (calculated_mos, noael_studies, ingredients, ghs_classifications, substance_identifiers, sensitization_assays, cir_safety_conclusions, dermal_absorption, cosing annexes II-VI, eu_clp_annex_vi) where (a) src_verification_status is in the verified set (sampled_verified, full_verified, salvaged_*, sampled_verified_external) AND (b) src_local_path is populated and points to an actual file on the build machine. Bulletproof = both conditions. Status alone is a guess; status + local_path is an address you can cd into. Out of 700,745 total fact rows the live count above is what passes both gates.
SUM(verified AND src_local_path NOT NULL) across 14 fact tables
No regulatory SaaS competitor publishes a binary source-traceability metric at this granularity — the count above is the highest-stringency provenance claim we can defend.
Live from benda-ingredients D1
700,764
Total fact rows across the 14 safety tables
Denominator for the bulletproof percentage. Sum of row counts across calculated_mos, noael_studies, ingredients, ghs_classifications, substance_identifiers, sensitization_assays, cir_safety_conclusions, dermal_absorption, cosing annexes II-VI, eu_clp_annex_vi.
SELECT SUM(COUNT(*)) across 14 fact tables
Live from benda-ingredients D1
106,951
NOAEL/MoS rows value cross-checked against ToxValDB or PDF source
Scope: calculated_mos (68,540 rows) + noael_studies (38,392 rows) = 106,932 tox-value-bearing rows eligible for cross-check. The count above is the subset where the stored NOAEL/MoS value has been confirmed within 2-10% tolerance against EPA ToxValDB ECHA IUCLID (71,242-row re-extraction of ECHA data) or PDF-verified against the SCCS opinion it claims to come from. Other tables (annexes, GHS, substance_identifiers, etc.) are authority-attributed but not NOAEL/MoS-bearing, so they are not eligible for this cross-check metric.
COUNT(*) FROM calculated_mos + noael_studies WHERE src_verification_status IN (full_verified, sampled_verified)
Live from benda-ingredients D1
30
Primary-source authorities in the registry
Canonical source_definitions table with url_pattern, doc_id_format, extraction_method, local_archive_path, retrieval_frequency, and qa_weight for each authority we attribute data to. Adding a new data source = one row in this registry.
SELECT COUNT(DISTINCT authority_short) FROM source_definitions
Live from benda-ingredients D1