How we build the supply chain graph

ForcedAlpha maps the chokepoints behind every AI, defense, and energy stock — and lets you audit each claim. Here's how.

The process

5 stages from raw data to published edge

01
Discover

We continuously scan public filings, press releases, government procurement records, and industry reports for new dependencies between companies, materials, and policy. Candidates enter a review queue and are never added to the live graph until they clear the following stages.

02
Evidence

Every relationship we add must be backed by a primary source — an SEC filing, official press release, earnings call transcript, or government record. We extract a verbatim quote and record the source URL. If no primary source exists, the edge is marked interpretive and held at a lower confidence tier.

03
Fact-check

Adversarial review agents verify every claim independently. Source URLs are fetched and the verbatim quote is matched against the live page content. Edges that fail quote-match or name-match are dropped entirely — not paraphrased, not downgraded. Every edge you see in the cascade viewer cleared this gate.

04
Score

Each chokepoint is scored on severity (1 to 5), evidence quality, and a structured rubric that covers the competitor landscape, substitutability, and policy exposure. The rubric is applied consistently across all nodes so scores are comparable across sectors and time.

05
Publish

New edges flow into our nightly graph rebuild and become queryable in the cascade viewer. Pro subscribers receive alerts when a new chokepoint clears a high quality threshold. Free users get the full audit layer — every edge has a source URL, confidence level, and a plain-language description of what would falsify it.

Coverage tiers

What we publish versus what stays internal

Public — free free
  • Bounded subgraphs around named chokepoints
  • Per-edge: primary source URL, verbatim quote, confidence level, and last-verified date
  • A plain-language statement of what would falsify each relationship
  • Coverage and freshness statistics updated nightly
Pro — $299/mo pro
  • Quality score alerts on new high-confidence chokepoints
  • Playbooks: chokepoint to ticker exposure to conviction score to entry and exit timing
  • 13F deep dives tracking institutional positioning around mapped nodes
  • Convergence dashboards and position-sizing tools
Boundaries

What we don't show, ever

We don't publish our scoring rubric weights or candidate-discovery pipeline. The value is what we choose to map and how we score it, not the underlying data. Publishing the implementation would let competitors replicate the methodology without the years of verification work behind it.
We don't publish unverified leads. If you see an edge in the cascade viewer, it has a source URL attached or it's explicitly marked interpretive. Edges still in the review queue are never visible to users.
We don't publish the full graph. The 11,000-edge map is the core asset. The public audit layer — source URLs, verbatim quotes, confidence levels, falsifiability statements — is what you should use to evaluate the quality of our work. It's a window into the methodology, not a substitute for it.
Current graph coverage
3,133
nodes mapped
11,329
edges
1,296
companies
See full coverage stats