Manufacturing.
Single-source nodes, certification cliffs, and the cascade nobody priced.
Modern manufacturing graphs are riddled with single-source nodes that don't show up on a balance sheet. A 5nm fab, a single-vendor photoresist, a sole-source rare-earth refiner — the system runs fine until one of them stops, and then it doesn't.
The hard part isn't enumerating suppliers. It's pricing the substitution gap: how long it actually takes to qualify an alternate part, recover capacity, and re-baseline the BOM. Most ERP and procurement systems can answer the first question and not the second.
When a node fails, the cascade isn't linear. Tier-2 vendors fall first, then a wave of contract amendments and force-majeure declarations, then the customer-facing impact. A reinsurer or a CFO needs to see that whole arc before the first headline lands.
Manifold treats the manufacturing graph as a directed causal network: foundries, materials, packaging, logistics, certification regimes, and end-products are all nodes; edges encode physical, contractual, or jurisdictional dependency.
Pillar scores light up the failure modes: I (Irreplaceability) catches the no-substitute nodes, R (Restoration Latency) prices the recovery clock, C (Cascade Load) measures the downstream footprint, T (Tail Depth) quantifies the bad case.
PEARL runs counterfactuals (what if this fab goes dark for 90 days?), PARETO simulates cascade dynamics, and SPIRTES discovers structure you didn't know existed in the production data.
Knows the named single-source risks. Doesn't have a tool that estimates the *recovery cost* of each one in dollars and weeks, or that surfaces the unnamed second-tier ones.
Per-node ΩF score, ranked by Cascade Load. Counterfactual recovery time on every critical path. A weekly report you can hand to the audit committee.
Operational risk is footnoted in the 10-K but not modeled. When a node fails, the loss is computed after the fact.
ΩSF (system fragility) and ΩSX (system exposure) as steady metrics. Tail-depth simulation puts a defensible number on the worst-case scenario, before it happens.
Pricing manufacturing-cascade exposure across portfolios is a black box. Most cat models stop at physical assets and ignore the network.
Counterfactual cascade simulation across the insured's full supplier graph. Defensible loss distribution that prices network risk, not just plant risk.
Needs to identify which nodes in the national manufacturing base are decisive — not which are biggest.
Irreplaceability + Cascade Load ranking on the national graph, configurable per strategic objective.
5nm capacity for AAPL, NVDA, AMD, QCOM. 37 downstream nodes fail with it.
Run Manifold on a manufacturing graph.
Trial accounts come pre-loaded with a curated dataset. Or request an invite to bring your own graph.
