What strata is, in one reading

Strata is a working platform, live now, that maps how critical minerals move through the world: which mines, plants and companies handle them, who ultimately controls those companies, and what the build-out will need, with the source of every fact attached. This page starts from what you can do with it and only then explains how it is built, in plain words with pictures. Every count on it is read from the same data the product serves.

You arrive with a name and a question

Most people come to strata holding a supplier, a plant or a mineral and one question about it: who controls this, how exposed are we, what happens if this link tightens. There are four ways to work that question. Each costs a different amount of your time, and every answer shows where it came from.

Look up any plant or company

Search covers everything the store holds, so a lookup costs a click and returns a link rather than a paragraph, with no model in the loop. These are the five kinds of thing it can find. The live counts are read from the shipped data when the page is built; the two graph figures are measured against the store on a stated date.

Commodities

the minerals themselves, each with a profile: production, reserves, trade, demand

69

profiles, read live

Facilities

mines, smelters, refineries and plants, each placed on the map and graded by evidence

504,276

commodity-attributed, read live · 942,330 in store on 2026-11-17

Companies

legal entities from registers and filings, reconciled so one company is one record

7,107,677

entity records · measured 2026-11-17

Ownership links

who holds whom, dated, with the percentage where a register discloses one

2,784,401

ownership and control edges · measured 2026-11-17

Trade flows

what crosses which border, by year, with export restrictions attached

every profile

by year, on each commodity's trade tab

Ask who really controls the plant

An operator's name is on the gate; the owner is usually somewhere else. Ask walks the chain of registered control from the plant upward and answers with the filings it used. Every link is dated, so a former parent stops counting when its filing stopped naming the company, and a minority stake, an index fund holding a few percent, is never treated as control.

From the plant to the ultimate parentoperated by100% ownedcontrolsended · not walkedminority · not a parentMinePeruOperating companyPeruHolding companyNetherlandsListed parentCanadaultimate parentFormer parentuntil 2020Index fund7% stake
walked: a control edge in forceended: a parent whose filing stopped naming this companyminority: a stake below control, never a parent
try it in Ask“Who ultimately controls the Cerro Verde copper mine, and through which companies?”/askcountry mapUnited States: who holds the mineral right, who operates, and who is only exploringExplore · Mines · United States

Measure the exposure, stage by stage

Explore ranks and compares minerals on concentration, foreign control and recycling gaps. Concentration is the headline supply-risk number, and it is not a black box: it is the sum of squared supply shares, computed separately at each chain stage, shown with the shares that build it. Nickel's mine stage is the worked example, and the comparison beside it is the whole point: a modest miner can be a dominant refiner.

The concentration bar on every screen is a single number - the Herfindahl-Hirschman Index (HHI): take each supplier's share of the total, square it, and add them up. It runs 0 (many equal suppliers) to 1 (one supplier owns everything). Squaring is the whole point - it makes a dominant producer count far more than several small ones. We show the shares, so you can rebuild the number yourself.

nickel Mine production2025
72%
Indonesia 72%Philippines 8%Russia 6%Canada 4%New Caledonia 4%
0.722+0.082+0.062+0.042+0.042+ …=0.54Highly concentrated

the top five shares alone already sum to 0.53; the dominant producer's square carries most of the index - which is exactly what a concentration measure should do

Diversified< 0.15Moderate0.15 - 0.25Concentrated0.25 - 0.50Highly concentrated≥ 0.50

The ladder runs green to amber and stops there - an index is a description, not an alarm, so there is no red. A missing bar means not enough suppliers disclosed to compute it, never "low".

HHI of which stage?

Because a metal's mine geography is not its refinery geography, we compute HHI separately at each stage and never blend them. The mine map and the refinery map are different maps - and the gap between them is the single most useful supply-risk signal on the page.

  • coppermine0.14Diversifiedrefining output0.20ModerateChina 44%
  • nickelmine0.54Highly concentratedrefining output0.28ConcentratedIndonesia 43%
  • lithiummine0.20Moderatechemicals output0.54Highly concentratedChina 70%

copper is the lesson: mine supply is the more diversified stage, but midstream routes through China - the risk lives downstream, and only a per-stage HHI shows it. "% attributed" on the live rows sizes how much of the total the named suppliers cover; a "Rest of world" residual is kept in production sums but excluded from the index, so it can neither dilute nor inflate the concentration number.

Project the demand, without pretending to own the model

Model runs a what-if. Strata does not run the world's climate and energy models and does not claim their science. It puts one conversation in front of 10 of them, serves PNNL's GCAM from 165 pre-solved worlds, and adds the layer none of them carries: the minerals the projected build-out will need, exported in the models' own input formats.

Ten models, one conversationone question/askGCAMEnergyHectorCarbonXanthosWaterTethysWaterDemeterLandMoiraiLandambrosiaFoodgaiaFoodMAWEINexusPrOMMiSProcessing
JGCRI / PNNLDOE, LBNL-led27 conversational tools · one contract: /ask /coverage /readyz

Trace every number to its source

Two kinds of number exist in strata. Some are reported by an authority; others are estimated from something countable multiplied by a verified coefficient. They look different everywhere they appear, and when the data cannot support a number the page shows nothing rather than a wrong one. The same honesty applies to the facility count below: a catalogued occurrence is not an operating plant, and the split says which is which.

Observed

Reported by an authoritative source - USGS, BGS, ICSG, a company filing, a national agency. The row carries its citation; click it and you land on the publication.

Estimated

Computed as observed activity × a verified material intensity. Always marked imputed, always shows BOTH ingredient sources, and expands to the full calculation on click.

The rule that holds the product together: a reader can always tell which of the two they are looking at - and when the data cannot support a number, we show nothing rather than a wrong number. Every observed row links to its publication; the full catalogue, ranked by authority, lives at /sources.

504,276 records

observed13,753
corroborated5,087
catalogued172,517
crosswalk152,794
occurrence160,125

Only the green share is load-bearing: 4% of records are observed directly or corroborated by an independent source. The rest is inventory: catalogued occurrences, crosswalked classifications and places a mineral was once found. Both are kept, and every page says which is which.

Why no single register could answer this

A wind turbine, an EV battery or a missile component is built from a few dozen minerals. Each is mined in one place, smelted in another, refined in a third and turned into a component in a fourth, by a different company at every step, and each step reports to a different register. The data exists; it was never written to be read together, and it fails in three specific ways.

Five hands, four countries, five kinds of recordmining cadastreMinePeru01company filingSmelterChina02industry surveyRefineryChina03customs codeComponentKorea04trade statisticsProductGermany05a different company at every step, a different register behind every company, and no register that sees the whole chain
Three hops between a plant and its real ownerplantholdingholdingstate ownerregistered locallycontrols itcountry statistics see only the first circle

Entity opacity

A refinery registered in one country can be controlled, three holding companies later, by a state enterprise in another. In every country-level statistic it looks independent.

You cannot see who really controls a plant from the country it sits in.

one customs codetwo markets
light elements~$100/kg
heavy elements~$1,000/kg

the trade statistic reports one number for both

Grade blindness

Customs codes bucket materials coarsely. One rare-earth code covers light elements at roughly $100 per kilogram and heavy ones at roughly $1,000. One number, two markets.

You cannot tell which industries are exposed from trade data alone.

Three methods, three islandsmethod Aown data?method Bown data?method Cown dataown data, own names: comparing two means rebuilding both

Method isolation

Every supply-risk method is built on its own data, its own company list and its own commodity names. Comparing two of them means rebuilding both from scratch.

You get one lens at a time, and no way to see where two disagree.

How it works underneath

For readers who want the machinery. Strata does one thing first: it reconciles 255 public and licensed sources into one canonical store, and everything above runs on that store rather than on its own copy of the data.

Many sources, one store, everything else on top◆Commodities▲Facilities●Companies⟋Ownership links⇄Trade flowsPeople askquestions, profiles, searchMachines callversioned files, model inputsMethods plug inscoring, demand estimates255 catalogued sourcesreconcileone store69 commodity profiles · provenance on every recordwhat runs on it

01

People ask

A question in plain language is planned into lookups over the store and answered with the rows and sources it used.

Ask

02

Machines call

Downstream models receive the same data as versioned files in their own input formats, never a screenshot of a page.

GCAM bundle manifest

03

Methods plug in

A supply-risk scoring or a demand estimate registers against the store and runs on every commodity it applies to.

How a method runs

A fact enters as the source wrote it, is reconciled against everything else the store knows, and only then becomes something a page or a model can use. Each layer can be rebuilt from the one beneath it, and the bottom one is never edited.

Raw, resolved, readyRawwhat the source saidUSGS row: Chile, 5,300 thousand tonnes, 2024Resolvedwhat we know after reconciliation+ commodity: copper · country: CL · source rank 1Readywhat the analytics produce+ share of world: 23% · concentration contributionone record, travelling upgains provenance at every layer

Raw

The file as it arrived, kept unchanged so every later step can be rebuilt from it.

Resolved

The same fact tagged with the canonical commodity, country and source rank, so it is comparable with every other fact.

Ready

Derived figures a page or a model can use, each pointing back to the rows that made it.

Fetchedparsed from a published source: which one, which URL, when, under what licenceCuratedasserted by a person, with the citation and the reasoning attachedDerivedcomputed from other records, carrying the chain back to each of them

Every record carries one of these three provenance kinds and a confidence tier, and a derived figure inherits the weakest confidence of anything that fed it. Each layer can be rebuilt from the one beneath it; the bottom layer is never edited.

Every register describes a site in its own words, and each is right. Reconciliation decides when several records are one asset, and more often proves that they are not. What a merge has to prove is on the reconciliation page.

before · 6 facility records

ICMMChuquicamata
USGS MRDSChuquicamata
USGS MASChuquicamata
WikidataChuquicamata
USGSChuquicamata Copper Smelter
USGSChuquicamata Copper Refinery

after · 3 facilities

Chuquicamata
one pit · 4 records merged
Chuquicamata Copper Smelter
kept separate
Chuquicamata Copper Refinery
kept separate

The smelter and refinery sit at the pit's exact coordinates and score as strong a match as the duplicates do. Proximity alone cannot tell them apart - the names can.

Three things strata is not, and three edges it keeps honest

The clearest way to say what something is, is to say what it is not.

Not a dashboard

A dashboard is one view of one dataset. Strata is the ground many views stand on, and the same store answers a question, renders a profile and feeds a model.

Not a risk score

Strata hosts scoring methods rather than championing one. Where several apply, each result is reported separately so the spread between them stays visible.

Not a model

The climate, energy and process models it serves belong to the groups that built them. Strata is data for those models, in their own input formats.

the honest edges

  • Coverage follows who publishes. A handful of countries release bulk cadastres, so the facility map is densest there and thinnest where no open register exists, and the capability brief measures that skew rather than letting it pass as geology.
  • Some of the vision is roadmap. What is built today is measured on the capability brief; what is aspirational is described as such and never rendered as if it were live.
  • A number is shown only when the evidence supports it. Where it does not, the page says so and shows nothing, which is a deliberate choice and the hardest one to keep.

Where to go from here

This page is the overview. Each of these goes one level deeper, in the same house style, and every one of them measures what it claims.

The source, commodity, facility, concentration and world figures are read from the shipped data each time the page is built. The company and ownership counts are the capability brief's measurements as of 2026-11-17. Where the data cannot support a number, the page says so rather than showing an estimate, the same rule the platform itself follows.