Methodology

How intelligence outputs are built and validated.

This platform is designed for traceable semiconductor research. Company, facility, country, and scenario views are grounded in structured records with explicit provenance, confidence, and freshness metadata.

1. What we track

Coverage model

We track companies, facilities, countries, production stages, materials, products, investments, logistics routes, chokepoints, and dependencies as linked entities. Aliases, parent-subsidiary relationships, and ownership structures are preserved so naming variance does not fragment the data model.

Sources are first-class records and can support multiple claims or relationships. Every major intelligence surface is intended to be read with source and freshness context, not as standalone narrative text.

2. Confidence scoring

Confidence model

Record confidence is normalized from five weighted factors: extraction confidence (0.30), source reliability tier (0.25), review state (0.20), cross-source agreement (0.15), and freshness (0.10).

Quality flags highlight weak-confidence records, low source coverage, stale evidence, duplicate candidates, and conflicting aliases. Flagged records are surfaced in analyst queues for review before broader publication.

3. Freshness policy

Freshness model

Freshness decays by time band and is shown across entity and operations views. Current scoring bands: up to 30 days, 90 days, 180 days, 365 days, and older than 365 days. Older records carry higher stale-risk weighting.

Source cadence schedules and refresh queues target stale records first. Enrichment follows a non-destructive policy: new evidence fills gaps and extends provenance without overwriting validated fields.

4. Reading outputs

Scores and scenarios

Scores are presented with weighted components and explanation text. Scenarios distinguish direct from indirect impacts and include confidence context so modeled propagation is visible rather than hidden.

Use this methodology page with each entity's Source panel for due diligence: the score or impact view should always be evaluated alongside source quality and freshness signals.

5. Bottleneck risk scoring

Three-axis country risk model

Country bottleneck scores are built from three independent axes — physical risk, supply concentration, and geopolitical risk — each scored 0–100 and combined into a composite. The composite is not an average: supply concentration is weighted most heavily because a sole-source dependency is structurally more fragile than a transient political risk.

Composite formula

Composite = 0.25 × Physical + 0.40 × Supply + 0.35 × Geopolitical

Supply concentration carries the highest weight (40%) because a country holding 80%+ of a critical input with no substitutes poses risk regardless of political or physical conditions. Geopolitical risk (35%) captures policy-driven disruption — export controls, sanctions, alliance realignment — which can materialize faster than physical events. Physical risk (25%) covers geographic and infrastructure hazards that are real but typically have longer warning windows.

Score thresholds

70–100 High — Active or near-term disruption risk. Requires diversification or contingency planning.
40–69 Medium — Structural vulnerability present; monitor closely, build buffer stock.
0–39 Low — Limited direct exposure; secondary or indirect risk only.

Physical risk axis — weight 25%

Geographic and infrastructure exposure

Physical risk scores capture hazards that could halt production or transit through non-political means. The primary factors:

Strait and chokepoint dependency — Sea cargo passing through the Taiwan Strait, Strait of Malacca, Strait of Hormuz, or Suez Canal faces potential blockade or piracy risk. Singapore and Malaysia score mid-range because nearly all their inputs transit the Malacca Strait.
Seismic and weather hazard — Taiwan, Japan, and the Philippines sit on active fault zones. The 2024 Hualien earthquake temporarily disrupted TSMC N3/N5 lines. Thailand's 2011 floods removed 25% of global HDD output for six months — the same industrial zones now host OSAT fabs.
Infrastructure reliability — Fab-grade water and power are not universal. South Africa's Eskom load-shedding disrupts mining smelters. Vietnam's power grid has caused yield events at Intel's Ho Chi Minh City campus.
Active conflict — Ukraine scores 95 (highest) because Cryoin and Ingas neon plants in Odessa/Mariupol were directly affected by the 2022 invasion. Israel scores 78 due to active regional conflict since 2023.

Supply concentration axis — weight 40%

Sole-source and market share risk

Supply concentration measures how much of a critical input depends on a single country, facility, or mine — and whether viable alternatives exist within a realistic ramp timeline.

Sole-source designation — ASML scores 98 because there is no other EUV lithography machine maker on earth. Soitec scores 80 because it is the only producer of SOI/FD-SOI wafers used by Intel, STMicro, and GlobalFoundries. ARM scores 88 because >95% of mobile processors use its ISA with no viable alternative.
Market share concentration — Countries controlling >60% of a stage score above 70. China controls >80% of gallium and >60% of germanium production; its 2023 export controls directly demonstrated this leverage.
Substitutability timeline — Even where alternatives theoretically exist, building a competing fab or mine typically takes 3–7 years. DRC cobalt scoring 85 reflects that while other deposits exist, refining infrastructure is entirely China-controlled.
Single-facility risk — ASML's entire EUV manufacturing occurs at one campus in Veldhoven. Carl Zeiss SMT's EUV optics come from a single site in Oberkochen. No backup exists for either.

Geopolitical risk axis — weight 35%

Policy-driven disruption risk

Geopolitical risk scores capture the probability that a government action — export controls, sanctions, alliance realignment, or domestic policy — disrupts supply. This axis moves faster than physical risk: a new Entity List designation can take effect within days.

Active export controls — China scores 92 because it is simultaneously subject to US BIS export controls on advanced chips and equipment, and has itself imposed gallium/germanium export restrictions. The Netherlands scores 68 because ASML now operates under a Dutch export license regime that the US directly pressured into existence.
Sanctions exposure — Russia scores 90: full US, EU, and UK sanctions since 2022 isolate its palladium exports and cut it off from Western chip imports. Hong Kong scores 85 because it has been treated as China under the US Export Administration Regulations since 2020, ending its special status as a separate customs territory.

Alliance alignment risk — Countries caught between competing US and China technology blocs score higher. Singapore and Malaysia navigate US export control enforcement pressure while maintaining trade relationships with China. The UAE scores 68 because it signed a US Technology Safeguard Agreement restricting AI chip use while simultaneously deepening technology ties with China.
Historical weaponization precedent — Countries that have previously weaponized supply carry elevated scores regardless of current political conditions. China's 2010 rare earth export halt to Japan and its 2023 gallium/germanium controls both establish a documented pattern of using mineral supply as geopolitical leverage.