Every company in our supply chain graph, grouped by theme. The grey dots are companies we scanned and rejected today. The glowing nodes passed a false convergence filter requiring high structural severity and top-quartile graph centrality. Click a glowing name to see the convergence narrative.
We built a supply chain graph with 2147 companies and 633 material nodes across 24418 dependency edges. Every edge represents a real supply relationship: mining, refining, supplying, fabricating, and integrating. Each company node carries a bottleneck severity score from 1 to 5 based on how substitutable it is, how geographically concentrated its supply is, and how long replacement would take.
The Heat Score Formula
Heat = (structural alpha) x (catalyst intensity) x (flow multiplier) / log(1 + market cap). Structural alpha comes from the graph itself: how many downstream dependencies does this node have, how severe is the bottleneck, how central is it to flywheel loops. Catalyst intensity is the sum of time-decayed, source-weighted data points from our 34 scanners. Flow multiplier captures directional tilt (more bullish activity pushes flow above 1.0, bearish below).
False Convergence Filter
Not every convergence matters. A senator buying MSFT generates activity, but MSFT is highly substitutable with multiple suppliers at every layer. Our filter requires high structural severity AND top-quartile graph centrality to reach the action zone. This rejects approximately 90% of all nodes daily. The rejection rate is the product.
Data Sources (34 scanners)
- Congressional trades, Senate eFDs, committee-relevant trade flagging
- 13F institutional holdings (24 managers), with new position detection
- DOD contract awards, with IDIQ tagging and work scope extraction
- SEC Form 4 insider transactions, 13D/13G activist stakes
- Options flow, IV spikes, CFTC positioning data
- Lobbying disclosures, corporate buyback execution, FTD/short interest
- Polymarket geopolitical probability spikes, credit stress indicators
Decay Rates
Each source has a half-life. Congressional trades decay in 14 days. DOD contracts persist for 30 days. 13F filings (which are already 45 days old when published) decay over 60 days. A data point below 5% of its original weight is dropped entirely. This ensures the heat map reflects what is happening now, not what happened three months ago.