AssayChain · Score Methodology · v2026-08-04
Every score is
auditable.
Formulas, inputs, weights, data sources, update cadence, and limitations for every heuristic score AssayChain produces. Nothing here is trade-secret — the formulas are simple, reproducible from the cited public data, and exist because verifiable data requires published formulas — not a black box.
1. Supply Risk Score
/api/risk/:commodity → supply_risk_score (0–100)
Purpose: Measure how vulnerable a mineral supply chain is to disruption at the country level.
Inputs (from data/risk-tiers.json, sourced from USGS MCS + CRFM):
top_producers: list of producing countrieschina_share_pct: China's share of global productionus_allied_share_pct: share from US + allied countries- Per-country
risk_score(1–5),tier("allied" / "adversarial" / "neutral"),alliedbool
Formula (simplified):
risk_score = base 40
+ china_share_pct × 0.3 (up to +30)
− us_allied_share_pct × 0.2 (reduces up to −20)
+ producer_country_avg_risk (average of top-producer tiers, 0–50)
clamp(0, 100)Data source: USGS Mineral Commodity Summaries (production shares), USTR/CRFM allied country lists.
Update cadence: Annual (tracking MCS 2025 baseline). Recompute when MCS 2026 publishes.
Limitations: Static annual snapshot — does not reflect real-time trade policy or shipping disruptions. Not an investment or procurement decision tool by itself.
2. Resilience Score
/api/resilience/:commodity → resilience_score (0–100)
Purpose: Stress-test whether a mineral supply chain could absorb a specific disruption.
Inputs: same risk-tier data + scenario parameters (supply cut %, duration).
Formula (5-factor weighted model, weights sum to 100%):
| Factor | Weight | Source |
|---|---|---|
| Allied supply ratio | 30% | risk-tiers.json us_allied_share_pct |
| Producer diversity (1/HHI) | 25% | USGS MCS top-3 shares |
| Domestic (US) production | 20% | USGS MCS US mine output |
| Stockpile/strategic reserve | 15% | USGS/DLA stockpile entries |
| Substitution potential | 10% | USGS substitution data |
Update cadence: Annual (MCS-driven). Scenario outputs recomputed on each call from the static baseline.
Limitations: Scenarios are illustrative stress tests, not forecasts.
3. Criticality Score
/api/criticality/:commodity → boolean flags + summary
Purpose: Multi-jurisdiction critical-mineral crosscheck. Not a numeric score — a membership check against the published lists.
Inputs: US DoD list, EU 2023 critical/strategic lists, IEA critical minerals list, Japan METI list.
Output: { us_dod: bool, eu_2023: bool, iea_2023: bool, jp_meti: bool } + count.
Update cadence: Quarterly review; lists change slowly.
4. Trade Controls
/api/trade-controls/:commodity → cn_export_restricted + notes
Purpose: China export-restriction status for Ga, Ge, graphite, Sb, W, Te, REE.
Inputs: PRC Commerce Ministry notices (2023–2025 restrictions), USGS MCS notes.
Output: { cn_export_restricted: bool, restriction_date, notes }.
Update cadence: On official notice; maintain a manual watch list.
5. Supply Concentration / HHI
/api/supply/:commodity → hhi
Purpose: Market concentration via the Herfindahl-Hirschman Index.
HHI = Σ (producer_share_pct_i)²- < 1500: Low concentration
- 1500–2500: Moderate
- > 2500: Highly concentrated (e.g., China REE)
Data source: USGS MCS top-producer shares (by % of global production).
Update cadence: Annual.
6. ESG Profile
/api/esg/:commodity → qualitative + quantitative fields
Purpose: Surfacing ESG-relevant factors per commodity.
Inputs (per commodity):
- Recycling share of supply (USGS/industry reports)
- Carbon-intensity by country of production (IEA/TDICI)
- Regulatory exposure (EU Battery Regulation, DFARS, IRA sections)
Output: structured per-commodity ESG fields; scores are qualitative tiers (low/medium/high) derived from the underlying documented values.
7. DLA Stockpile Status
/api/stockpile/:commodity → status + quantity
Purpose: Reported US strategic stockpile levels.
Inputs: USGS MCS "stockpile" table (published annually), DLA public filings.
Output: { status: string, quantity, unit } where available.
Limitation: DLA rarely publishes live levels; these are annual best-available figures.
8. Compliance Check / NI 43-101 / JORC assistive only
Important: These endpoints are AI-assistive screening tools, not legal, financial, or regulatory advice. Every response carries a warning field stating they are NOT a substitute for Qualified Person (QP) / Competent Person (CP) review.
How checklists are built:
- NI 43-101: 8 checklist items (QP sign-off, QA/QC, chain of custody, data verification, lab accreditation, independent review, disclosure standards, report format) mirrored from NI 43-101 / Companion Policy 43-101CP.
- JORC 2012: 10 Table 1 items across sampling/tenure/reporting sections, mirrored from the JORC Code 2012 Edition.
- Status values:
pass/needs_review/failderived from (a) country-region compliance flags indata/compliance.json, (b) optional caller-suppliedcompetent_personfield, (c) default toneeds_reviewwhenever a document-level judgment is required.
Limitation: No LLM-generated checklist qualifies as QP/CP sign-off. None of these endpoints prepare a filing-ready report.
9. Ore Potential
/api/ore-potential → preliminary_assessment
Purpose: First-pass in-situ value + qualitative deposit-type classification.
Inputs: Caller-supplied assay rows + attested USGS benchmark prices + surrounding district data.
Output: estimated_in_situ_value_usd_per_tonne, deposit_type, confidence, recommended_tests.
Limitation: Preliminary screening for triage only — requires follow-up lab work and a geologist's interpretation for any project decision.
10. Index / Data Source Master List
| Dataset | Source | License |
|---|---|---|
| Commodity production/reserves | USGS Mineral Commodity Summaries | Public domain |
| District histories | USGS MRDS / state surveys | Public domain |
| Price snapshots | LBMA, FRED, Yahoo Finance | Redistribution: see source terms |
| BLM claims | BLM LR2000 | Public domain |
| Earth MRI focus areas | USGS | Public domain |
| Risk tiers | USGS MCS + USTR/CRFM lists | Public domain / derived |
| Compliance lists | NI 43-101, JORC, DFARS, IRA, EU Reg 2023/1542 | Public regulations |
Known Limitations (honest)
- No real-time data. Annual/quarterly static baselines drive most scores.
- Heuristic, not regulator-endorsed. These formulas are the operator's design.
- Country-level only. No deposit-level or company-level supply-chain visibility in the core risk endpoints.
- Compliance = assistive. Always requires QP/CP review before any filing use.
- Extraction quality grading (A–D) is a strict quality-gate heuristic with 80% refund on Grade D — it is not a guarantee of correctness on any single document.