Methodology

Purpose

lanthanides.io tracks real-world prices for rare earths, rare metals (tantalum, niobium, cobalt, lithium), strategic metals, and semiconductor materials. It is designed for procurement analysis, supply-risk monitoring, and market intelligence.

All displayed prices require source provenance. No data is fabricated, interpolated, or synthetically generated.

---

Display Prices

Each element page shows two headline prices representing distinct market tiers.

Retail Reference

The retail reference is the lowest verified, in-stock price from an established Western retailer for a practical quantity (roughly 10 to 100 g). It answers the question: what would it cost to buy this element today?

Selection criteria:

• Source: Established retailers with public storefronts and track records. Anonymous or unverified marketplace listings are excluded.
• Stock: Must be currently available, not sold out, backordered, or "notify me" status.
• Form: Metal is preferred when available. Where only oxide or other forms are sold at retail, the form is stated explicitly.
• Quantity band: The listing in the 10 to 100 g range yielding the lowest normalised cost (USD/kg) is selected. Listings under 5 g are excluded to avoid small-quantity premiums that distort the rate. Listings over 1 kg are excluded as wholesale.
• Purity: The most commonly available commercial purity, generally 99.9% to 99.99% for rare earths.
• Confidence: The selected record must have a confidence score ≥ 0.5.

The retail reference is a single observed price, not a computed average. It is directly actionable. A reader can visit the cited seller and purchase at or near this price.

Bulk Benchmark

The bulk benchmark is the most recent credible price for industrial-scale quantities of the element's primary trade form, usually oxide for rare earths. It answers the question: what is this material worth at commodity scale?

Source hierarchy (in order of preference):

1. Recognised price-reporting agency (Fastmarkets, Argus, Shanghai Metals Market, IMARC)
2. USGS Mineral Commodity Summary unit value (if published within 12 months)
3. Corroborated direct-seller quote with MOQ ≥ 10 kg and confidence ≥ 0.6
4. Procurement index (BusinessAnalytiq or equivalent), labelled as index-derived

Delivery basis is stated with each benchmark (e.g., EXW China, FOB China, CIP NE Asia). Where no commodity benchmark exists, the field displays "n/a" with an explanatory note.

Retail Premium

Where both prices are available, the ratio Retail Reference ÷ Bulk Benchmark is displayed. This ratio quantifies the combined cost of reduction, refining, packaging, certification, and dealer margin above the commodity input price. For rare earth metals, this ratio typically ranges from 2× (abundant light rare earths) to 100×+ (ultra-pure research grades of scarce elements).

Why not an average?

Previous versions of this site displayed a confidence-weighted mean across all retail records. This was retired because averaging across structurally different market segments (collector specimens, bullion bars, laboratory reagents, investment platforms) produces a number that corresponds to no real transaction. A PEGUYS 10 g vial, an RWMM bullion ingot, and a Sigma-Aldrich 1 g reagent jar are not three noisy observations of the same price. They are three different products.

Commodity price benchmarks (Fastmarkets, Argus, SMM) solve this by defining a single reference product with fixed specifications and surveying transaction prices for that product only. The two-reference approach adopted here applies the same logic: each displayed number refers to one specific product class, clearly identified.

The full offers table on each element page preserves all collected data for readers who need the complete picture.

---

Price Tiers

Retail

Small-quantity purchases available to individuals, laboratories, or small businesses. Sold by online retailers, specialty chemical suppliers, or small distributors. Quantities typically under 25 kg.

Bulk / Industrial

Larger-quantity purchases through industrial distributors, manufacturers, or commodity brokers. MOQ typically ≥ 25 kg. Often require direct quotation with negotiated terms.

Lab-grade

Research-quantity purchases from scientific suppliers (Alfa Aesar, Sigma-Aldrich, Chemsavers, MSE Supplies). Per-gram costs are often 10 to 1,000× above commodity prices due to certification, packaging, and ultra-high purity. These records are included in the offers table for completeness but are never selected as the retail reference due to their extreme per-kilogram rates.

Retail and bulk prices are never merged or averaged. They represent different markets.

---

Normalisation

All prices are normalised to USD per kilogram for comparability.

• Currency: Converted at the exchange rate recorded when the quote was collected. Rate and date are stored with each record.
• Units: Per-gram, per-pound, per-ounce, or per-tonne prices are converted using standard factors.
• Original preservation: The original quoted price, currency, and unit are always retained. The normalised value is a derived field.

Normalisation never changes the underlying quote. Always refer to the original values in the provenance table for precision.

---

Oxide-to-Metal Conversion

Most rare earths are primarily traded in oxide form. When comparing oxide and metal prices, the stoichiometric relationship and processing premium must be considered.

Stoichiometric mass fraction

For a rare earth oxide RE_xO_y, the metal content is:

f_RE = (x · A_RE) / (x · A_RE + y · A_O)

where A_RE is the atomic mass of the rare earth and A_O = 15.999 u.

Oxide	Formula	Metal fraction	Metal from 1 kg oxide
Scandium oxide	Sc₂O₃	0.652	652 g
Lanthanum oxide	La₂O₃	0.853	853 g
Cerium oxide	CeO₂	0.814	814 g
Neodymium oxide	Nd₂O₃	0.857	857 g
Terbium oxide	Tb₄O₇	0.850	850 g
Dysprosium oxide	Dy₂O₃	0.871	871 g
Holmium oxide	Ho₂O₃	0.873	873 g

Metal price estimate from oxide

P_metal,est = (P_oxide / f_RE) × m

where m is the metal-over-oxide multiple, typically 1.5 to 3.0× for rare earths, reflecting reduction costs (calcium or lithium reduction under vacuum or inert atmosphere).

This conversion is for analytical reference only. Display prices are taken directly from observed records and do not apply stoichiometric conversion.

---

Material Forms and Purity

Prices are tracked separately by form (oxide, metal, powder, alloy, compound, etc.) and by purity (99%, 99.5%, 99.9%, 99.99%, etc.).

The retail reference price always states its form and purity. No aggregation across forms or purities occurs in the headline display.

---

Verification and Confidence

Verification statuses

Status	Meaning
Verified Invoice	Extracted from an uploaded invoice with quantity, date, price
Corroborated	≥ 2 recent independent offers aligned on price range
Single-Source Offer	One recent public offer; no independent confirmation
Benchmark-Linked	Supported by a recognised index or benchmark entry
Stale	Older than 90 days; directionally useful but not current
Archived	Hidden from current views; retained for history

Confidence scores

Range	Interpretation
0.8 to 1.0	High: invoice-backed or well-corroborated
0.5 to 0.79	Medium: plausible, limited corroboration
0.0 to 0.49	Low: single source, unclear provenance

Confidence is assigned during ingestion based on source type, parse quality, corroboration, and recency. The confidence score determines eligibility for reference-price selection (minimum 0.5 for retail reference, 0.6 for bulk benchmark) but does not affect a weighted average. There is no weighted average.

---

Source Categories

Price data is collected from four principal source categories.

Specialty element retailers

Established Western retailers with public storefronts selling individual elements in collector, educational, and laboratory quantities. Includes dedicated element shops, eBay sellers with multi-year track records and verifiable feedback histories, and bullion dealers. These provide retail reference prices.

Chemical and laboratory suppliers

Companies selling reagent-grade, research-grade, or industrial chemical products. Includes both large catalogues (Sigma-Aldrich, Alfa Aesar, MSE Supplies) and smaller specialty chemical distributors (Chemsavers). These records are included in offers tables for completeness but are generally not selected as the retail reference due to extreme per-kilogram rates at laboratory scale.

Chinese trading platforms and manufacturers

Alibaba-verified suppliers, direct manufacturer quotes, and Chinese commodity indices. These sources provide bulk benchmark pricing. For export-controlled elements, listed prices may not reflect obtainable export pricing. The applicable regulatory status is always noted alongside.

Direct supplier quotes and invoices

Quotations solicited directly from distributors, manufacturers, or trading companies, plus uploaded purchase invoices. These carry the highest trust tier and are the preferred source for both retail and bulk reference prices.

Solicited supplier quotes are collected through a consent-gated workflow. A supplier appears in the outreach registry only after explicit opt-in (consent.opted_in: true), with a dated note recording where the consent came from, and contact is rate-limited per supplier. When a supplier replies with a price, the maintainer enters the structured fields through outreach/intake.py, which:

• validates the input (rejecting non-positive prices, future dates, unknown tiers, or missing required fields),
• converts non-USD prices using a manually maintained FX table at outreach/fx.yml; no live FX API is called, and a currency missing from that table is refused rather than guessed,
• preserves the original quoted price, original currency, and FX rate used alongside the USD-normalised value,
• appends a dated observation to the element's price history with source_type: supplier_quote and the supplier id as the source,
• and records the contact in the registry so the per-supplier rate limit is honoured on the next outreach run.

The source_type tag lets the front end and downstream analysis distinguish maintainer-collected quotes from public-listing prices. If a supplier omits a value (purity, form, exchange-rate context), the corresponding field is left blank rather than guessed. No prices are invented. Suppliers can withdraw consent at any time and are removed from outreach immediately; the observations they previously contributed remain in place, attributed to them by id and dated as of the original reply.

Permanently excluded sources

Certain sources are permanently excluded from the database regardless of apparent pricing. Exclusion criteria include: sellers with documented patterns of misrepresenting purity or origin; platforms where provenance cannot be independently established; aggregator sites that republish prices without source attribution; and any source where repeated verification attempts have produced inconsistent or fabricated data. Excluded sources are not named publicly but are tracked internally.

---

Provenance Chain

Every price observation in the database arrives through one of three intake paths. All three apply the same validation rules (positive price, non-future date, known tier, known element, FX rate present for non-USD) and every observation carries an explicit source_type tag identifying which path it came from. No observation is ever fabricated, interpolated, or synthesised from a model.

Public listings (source_type: public_listing)

Prices observed on a public storefront: specialty retailers, eBay sellers with verifiable feedback histories, Alibaba pages, lab-supply catalogues. Collected by the maintainer or the data pipeline with the seller name, date, form, purity, quantity, and (where available) the original URL recorded alongside the price. A reader can, in principle, visit the cited URL and verify the listing.

Community submissions (source_type: community_submission)

Prices submitted by readers through the price-update GitHub issue template. The path has two human checkpoints by design:

1. The maintainer reads the issue. Until the maintainer applies the approved label, no automation will touch the data. The issue might describe a real and useful observation; it might also describe something stale, mis-typed, or speculative. The maintainer is the one who decides the difference and records that decision by labelling the issue.
2. A pull request opens with the data change. The maintainer-triggered workflow at .github/workflows/community-intake.yml (manual workflow_dispatch only, never automatic on issue creation) parses the approved issue, validates the fields with the same rules used for listings and quotes, and opens a PR adding one observation to _data/price_history/<symbol>.yml. Merging the PR is what publishes the observation. Closing it without merging discards the submission cleanly. The submission carries the original issue number for traceability.

A submission that cannot be validated (bad price string, missing field, future date, unknown currency) fails loudly in the workflow run rather than silently producing a half-record. There is no path from an unreviewed, unlabelled, or malformed issue into the price history.

Supplier quotes (source_type: supplier_quote)

Prices replied directly by suppliers who have opted in to the outreach registry. The opt-in is recorded with a date and a short note describing where consent came from; a supplier without consent.opted_in: true is never contacted and never has a quote recorded against them. When a supplier replies, the maintainer reads the message and runs outreach/intake.py to enter the structured fields by hand: no scraping, no auto-parsing of free-form text, no live FX API. The original quoted price, original currency, and FX rate used are preserved alongside the USD-normalised value, and the contact is recorded in the registry so the per-supplier rate limit is honoured on the next outreach run. Suppliers can withdraw consent at any time; the observations they previously contributed remain in place, attributed to them and dated as of the original reply.

Shared guarantees across all three paths

• Validated. Every observation goes through scripts/price_history.py's validator: positive price_per_kg, non-future date, known tier (retail, bulk, lab), known source_type, non-empty source.
• Dated. Every observation carries the date it was observed, quoted, or invoiced, never an ingestion date dressed up as a quote date.
• Attributed. Every observation names its source. For listings, that's the seller and (where available) URL. For quotes, the supplier id from the registry. For community submissions, the seller / URL the submitter gave, plus the issue number.
• Never invented. No field is filled in to "complete the record". If a supplier omits purity, the purity field is blank. If a community submission omits a date, the workflow refuses to write the row.

---

Data sources breakdown

The mix of observations currently in the database, by intake path. Regenerated by scripts/source_breakdown.py from _data/price_history/*.yml. This table reflects what the data actually contains, not an aspirational target.