apex-7 · layer 4 of MarginOps

Decisions, actions, financial outcomes, with confidence on every edge.

This is the differentiator that competitors do not have. Each historical decision is stored as a node. Each downstream action is connected with an edge weight. The terminal outcome resolves in dollars from gold.fact_daily_revenue.

nodes · brand

1,284

decisions · actions · outcomes

edges

3,940

weighted · directional

avg confidence

0.74

trailing 90d

Causal chain · "Cut Thursday lunch FOH by 1"

drilling 1 decision

Each edge weight is the conditional probability that this action followed from the prior decision, learned from 90 days of operator history. Outcome dollars resolve from gold.fact_daily_revenue diffed against the baseline window. The graph self-prunes weak edges below w < 0.25 nightly.

node schema

CREATE TABLE platinum.intelligence_nodes (
  node_id     UUID PRIMARY KEY,
  org_id      UUID NOT NULL,
  node_type   TEXT NOT NULL,
  name        TEXT NOT NULL,
  confidence  NUMERIC(3,2) NOT NULL,
  detected_at TIMESTAMPTZ NOT NULL
);

edge schema

CREATE TABLE platinum.causal_chains (
  chain_id     UUID PRIMARY KEY,
  source_id    UUID NOT NULL,
  target_id    UUID NOT NULL,
  weight       NUMERIC(3,2) NOT NULL,
  financial_impact_cents BIGINT,
  window_start DATE NOT NULL,
  window_end   DATE NOT NULL
);

why this matters

Restaurant365 and MarginEdge surface metrics. They do not learn the dependency.
This graph answers "what happened the last 4 times I tried this?"
The investor-deck demo lives here. Pull one decision, show the chain.
Recommendations in the morning brief cite specific node IDs.