Supply chain disruption on a typed graph

This is an anonymized design-partner scenario; the institution is a manufacturer, the figures are illustrative, and no actor has agreed to anything. We are pre-revenue, building the POC. With that label on it, here is what a supply-chain shock looks like when the substrate is a four-layer typed graph, not a flat feed.

A flat feed knows the symptom

A flat feed says "the part is short." It does not say why. The shortage could be a supplier outage (L3 Company), a port shutdown (L1/L3), a tariff (L2 Government), or a strike (L4 Actor, public signals only). All four look the same in a flat feed — the part is short — and the response is the same for all four, which is the problem. The GNL splits the explanation across layers, so the cause is a first-class, typed, versioned thing, not an edge to a single node.

Which layer drove which move

The GNL lets a query ask which layer drove which move. The tariff is an L2 NodeVersion; the supplier rerouting is an L3 NodeVersion; the contract repricing is an L1 NodeVersion; the port-worker action is an L4 NodeVersion, public signals only. The shock is not one event; it is a cascade across layers, and the typed graph makes the cascade a path a planner can read. The flat feed knows things moved; the typed graph knows which layer made them move — and that is the difference between a symptom and an explanation.

A flat feed knows the part is short. A typed graph knows which layer made it short — and that is the difference between a symptom and an explanation.

• L1 Instruments (contracts), L2 Governments (tariffs), L3 Companies (suppliers), L4 Actors (public signals) — the four layers of a shock.
• A cascade is a path across layers, not a single edge — the typed graph makes it readable.
• L4 is public signals only, by design — the strictest governance layer is a choice, not a default.

The manifest that explains the reroute

A planner that reroutes a part on a flat feed is rerouting on a symptom. A planner that reroutes on the GNL is rerouting on a typed explanation, and the manifest records which NodeVersions the reroute read — so the decision is reproducible. A regulator or auditor asking "why did you reroute?" gets the manifest, not a narrative: the L2 tariff version, the L3 supplier version, the L1 contract version that drove the call. The manufacturer does not get a certificate that the reroute was optimal; they get a defensible record of why the shock was explained the way it was. Calibration of any forecast over the graph is gated (Gate 1b); the manifest is the audit story, not the accuracy story.

A flat feed says a part is short. A typed graph says which layer made it short, and the manifest records the versions the reroute read — the symptom becomes an explanation, and the explanation is reproducible.

— design-partner supply-chain note