Project TBD Memory Vault

2026-04-15-data-source-strategy

5 min read

Data Source Strategy

Semiconductor Supply Chain Ontology — The Data Landscape

Ross & Perry · Stanford GSB · April 2026

A priority-ranked map of 47 sources to seed the digital twin — what’s free, what’s gated, what to hit first.

47 Total Sources | 28 Free/GSB Access | 15 P1 Must-Haves | 0 P1 Sources Behind Paywall

How to Read This Deck

Access Classification

  • FREE — Publicly accessible. No credential.

  • GSB — Available via Stanford GSB library (Bloomberg, Capital IQ, Panjiva, FactSet, etc.).

  • PAID — Commercial license required. Not GSB-accessible.

  • VARIES — Tiered: some data free, full datasets require membership.

Priority Tiers

  • P1 MUST HAVE — Core to ontology population. Day-1 ingestion targets.

  • P2 HIGH VALUE — Enriches the graph: product-to-facility linkages, ownership trees, trade flows. Phase 2.

  • P3 ENRICHMENT — Dynamic updates, facility verification, event monitoring. After core graph established.

Thesis: The free + GSB tier is sufficient to seed the ontology skeleton. Paid sources buy velocity via pre-resolved relationships — defer until a design partner or post-prototype.

The Matrix: Priority × Access

P1 · MUST HAVE

  • FREE (11 sources): EDGAR, TWSE, KRX, Euronext, BIS, OFAC, UFLPA, USITC, Census, Comtrade, CHIPS.gov

  • GSB (2 sources): Bloomberg SPLC, Capital IQ (Corp Trees)

  • PAID (0 sources): None

Every P1 source is free or GSB-accessible. Zero P1 dollar-cost.

P2 · HIGH VALUE

  • FREE (13 sources): Company IR, ESG, Form SD, DOD 1260H, ITC 337, METI, EU Dual-Use, Taiwan MOEA, Korea MOTIE, Eurostat, Taiwan Customs, SIA, Patents

  • GSB (6 sources): Panjiva, Refinitiv, FactSet, Gartner, IDC, PatSnap

  • PAID (6 sources): Omdia, Yole, TechInsights, IC Insights, Orbis, ImportGenius

P3 · ENRICHMENT

  • FREE (5 sources): WTO TPR, OSM/Google Maps, EPA, GDELT, SemiAnalysis

  • GSB (1 source): Factiva

  • PAID (2 sources): Planet Labs/Maxar, DigiTimes

The 10 Source Categories

  1. Corporate Disclosures — Supplier dependency, geographic risk, subsidiary trees. EDGAR is the richest free source. → Company, Facility nodes

  2. Regulatory & Gov Lists — Compliance layer. BIS, OFAC, UFLPA — free, machine-readable, updated daily. → SubjectToRegulation edges

  3. Trade & Customs — HS 8541/8542/8486 bilateral flows. Comtrade free; Panjiva adds shipper-consignee pairs. → ShipsProduct edges

  4. Market Intelligence — Pre-resolved relationships, market share, capacity. Gartner/IDC via GSB; Omdia paid. → Ownership, market share weights

  5. Industry Bodies — SEMI fab census, SIA factbook, JEDEC standards. → Product taxonomy

  6. Geospatial / Facility — OSM + CHIPS.gov free; Planet Labs for construction tracking (defer). → Facility verification

  7. Patents & Technology — Assignee mapping reveals R&D footprint and IP dependency (ARM, RISC-V). → TechnologyDependsOn edges

  8. Financial Infra — Bloomberg SPLC, Capital IQ — pre-resolved corporate hierarchies. → Entity resolution backbone

  9. Academic Research — RAND, CSIS, CNAS on geopolitical risk; NBER for empirical supply chain models. → Risk overlays, validation

  10. News & Events — GDELT structured events; DigiTimes for Taiwan signal; SemiAnalysis for deep technical. → Dynamic disruption signals

P1 — Must Have (Week 1 Targets)

Every P1 node is reachable with zero commercial spend. Entity resolution (EDGAR ↔ Bloomberg ↔ Cap IQ) is the bridging skill.

P2 & P3 — Enrichment Layers

P2 · HIGH VALUE

  • Panjiva (GSB) — Bill of lading: shipper/consignee pairs EDGAR doesn’t disclose

  • Gartner + IDC (GSB) — Market share by segment (fabless/foundry/OSAT/IDM)

  • Omdia / Wood Mac (Paid) — Fab-level capacity, capex tracking, node roadmaps

  • Yole Group (Paid) — Advanced packaging supply chains — buy 2-3 targeted reports

  • TechInsights (Paid) — Die-level supplier ID via teardowns — expensive, target key chips

  • USPTO / EPO Patents — IP dependency mapping (ARM licensees, RISC-V adopters)

  • Refinitiv · FactSet (GSB) — Geographic revenue, supply chain module

  • DOD 1260H · ITC 337 — Litigation-surfaced relationships, Chinese military firms

P3 · ENRICHMENT

  • GDELT (Free) — Structured event extraction from global news — bulk BigQuery

  • SemiAnalysis (Free) — Deep technical supply chain analysis — subscribe free tier

  • DigiTimes (-3K/yr) — Highest signal-to-noise for TSMC/Taiwan supply chain

  • Factiva (GSB) — Historical archive — use for event back-fill

  • OSM + Google Earth — Fab/OSAT location verification — Overpass API for bulk

  • EPA Permits (Free) — U.S. facility confirmation + capacity correlation

  • Planet Labs / Maxar — Satellite construction tracking — defer to post-Series A

  • RAND / CSIS / CNAS — Geopolitical risk overlays — free high-quality reports

Paid-spend posture: Buy 2-3 targeted Yole reports + DigiTimes sub now (~-15K). Defer Omdia, TechInsights, Planet Labs until design-partner or post-prototype.

Week 1 Acquisition Sequence

  1. Day 1 — Regulatory Lists: Download BIS, OFAC, UFLPA. Merge into unified restricted-party table. Compliance layer live.

  2. Day 2 — EDGAR Bulk: Run sec-edgar-downloader on top 20 semi tickers. LLM-extract supplier mentions from 10-K/20-F.

  3. Day 3 — Trade Flows: UN Comtrade + USITC DataWeb pulls for HS 8541/8542/8486, 2019-2023. Seed ShipsProduct edges.

  4. Day 4 — GSB Stack: Bloomberg SPLC for top 20 tickers. Capital IQ corporate tree exports. Resolve entity aliases.

  5. Day 5 — Entity Resolution: Build master alias table. LEI as primary key via OpenCorporates. OpenRefine / recordlinkage for fuzzy matching.

Outcome: ~200 company nodes · 500+ relationship edges · full compliance layer operational · entity resolution backbone in place

  1. Audit GSB access now — Email the library: confirm Bloomberg SPLC, Capital IQ bulk export, Panjiva, Refinitiv, FactSet, Gartner/IDC. Bulk-export permissions vary.

  2. Run Week 1 sequence — Regulatory lists → EDGAR → Comtrade → GSB stack → entity resolution. Five days to a live ontology skeleton.

  3. Entity resolution is the unlock — TSMC appears as 5+ aliases across sources. Stand up LEI-keyed master alias table early.

  4. Defer paid spend, buy targeted — Skip Omdia/TechInsights/Planet Labs for prototype. Pull trigger on ~2 Yole reports + DigiTimes sub when Phase 2 kicks off.

  5. Convert table to Cowork tracker — Use the prompt in the source doc to generate a live research tracker spreadsheet.

Source: ProjectTBD_DataSource_Strategy.pptx.pdf — also stored in Obsidian vault at Project-TBD/Research/ProjectTBD_DataSource_Strategy.pdf

Graph View