Methodology
How the tracker decides what it knows.
The product separates sources from claims, labels every public claim with confidence and evidence status, and treats workforce and operating evidence as first-class fields rather than footnotes.
01 Thesis
The issue is downstream capacity.
Financing, diplomacy, stockpiles, price floors, and demand guarantees are necessary instruments. They are not the same thing as operating mines, separation trains, alloy lines, powder plants, and sintered-magnet furnaces. For each instrument, the tracker asks whether the capabilities it would need are present, and how the workforce behind those capabilities is being built.
The product’s analytical lens is capability conversion: the gap between announcing an instrument and producing the physical output it is supposed to enable.
Talent is treated as embedded in capability, not a parallel column. Solvent-extraction trains and sintering furnaces are run by people whose judgment is the operating asset.
Tacit process know-how, plant commissioning, customer qualification, and environmental and permitting capacity are given equal weight with capital and policy.
02 Source hierarchy
Where each kind of source sits.
Sources are classified by type and assigned a confidence label. The classification governs how a source can be used: signal sources, for example, can mark leads but cannot stand alone as the basis for a public claim.
Official
Government, regulator, and intergovernmental publications. High trust for facts the issuer directly publishes; interpretation by the issuer is treated as a source claim.
Institutional data
Recurring datasets from established institutions (e.g. USGS, IEA, multilateral banks). High trust for aggregated data; cite the most recent edition.
Think tank
Policy research institutions with editorial standards. Strong for framing and synthesis; treat specific quantitative claims as secondary unless cross-checked.
Company
Company filings, press releases, investor materials. High trust for self-disclosed historical facts; forward-looking claims are treated as company positions.
News
Established outlets reporting facts and statements. Cross-check against primary sources before using as the only support for a public claim.
Academic
Peer-reviewed research and university working papers. High trust for methods and historical analysis; expect publication lag relative to live events.
Professional association
Industry bodies (e.g. SME, SAE, magnet-industry associations). Useful for industry norms, qualification practices, and labor-market context.
Signal
OSINT, social, conference chatter, analyst commentary, and working-name references. Signal-only — never the sole basis for a public claim.
03 Claim and evidence model
Sources are separated from claims, and claims carry evidence labels.
A source is a document, dataset, or signal. A claim is a specific fact, estimate, interpretation, forecast, risk, or lead supported by one or more sources. Every public claim carries a confidence label and an evidence status. The two are not the same: a claim can be high-confidence but only partially supported, and vice versa.
Confidence answers: how much should a reader weight this claim when forming a view? Evidence status answers: what kind of support stands behind it?
A primary-confirmed fact in an authoritative dataset is typically high confidence. A widely cited industry estimate that is consistent across analysts is typically medium confidence and secondary-supported. An unverified-lead claim is low confidence and is held to be tested, not quoted.
Claims may be tagged as facts, estimates, interpretations, forecasts, risks, or leads. The type signals how the claim should be read — an interpretation is editorial, a lead is provisional.
04 Confidence labels
What high, medium, and low mean.
- Directly supported by one or more primary or institutional sources with no material contradicting evidence on record.
- Supported by reliable secondary sources, or by primary sources with an interpretation gap that the editorial team has bridged in good faith.
- Working hypothesis, framing concept, or partially cited entry. Held in the tracker to be tested, not to be quoted.
05 Workforce mechanism
What it means to say a policy has a workforce mechanism.
Most public commentary on rare-earth policy treats workforce as a footnote. The tracker treats it as a first-class field. For each assessment, the workforce mechanism is rated by how concretely it is tied to the announced instrument.
- No workforce instrument identified inside or alongside the initiative.
- Workforce considerations are referenced but not designed into the instrument; outcomes depend on actors outside the program.
- A workforce mechanism is articulated as part of the initiative, with a stated theory of change but limited or unspecified funding.
- A workforce mechanism with dedicated funding and a delivery vehicle (grants, apprenticeships, training partners).
- A workforce mechanism that is actively training, placing, or retaining workers with measurable throughput.
06 Operating evidence
From announcement to independent confirmation.
An initiative can be announced, financed, signed, piloted, scaled, and externally confirmed. Each step is a different epistemic state. The tracker labels each assessment with the highest step for which evidence exists.
- Announced or rumored only. No operating activity tied to the initiative.
- Program announced, instrument signed, or counterparty named. Capacity build has not yet begun.
- Pilot- or demonstration-scale activity is underway against the announced outcome.
- Commercial-scale activity is underway as a direct result of the initiative.
- Commercial-scale activity is independently confirmed by parties who are not sponsors of the initiative.
07 Limitations
What the tracker is not.
The tracker is deliberately narrow and deliberately editorial. Reading it well means reading these limits.
- 01Current scope is deliberately narrow: rare earths and NdFeB permanent magnets. Other critical minerals are out of scope until this model proves useful in this column.
- 02The data is static and editorial. There is no scraping, no live pipeline, and no API. Each pass is a hand-curated snapshot.
- 03The seed source register is small. Several entries now have official public sources, but confidence labels still distinguish official existence from operating-capacity evidence.
- 04Pax Silica, Project Vault, FORGE, and the DOE Career Map now have official public-source anchors. Their capacity-conversion impact remains a separate assessment.
- 05Severity, transferability, risk, and operating evidence are qualitative judgments. Numerical scores would imply precision the underlying evidence does not support.
- 06The tracker covers institution-, role-, company-, and public-official level data only. No private individuals are tracked.