Building the graph

To demonstrate how this data model enables intelligent features, I built a working Neo4j knowledge graph populated with realistic synthetic data representing the private equity ecosystem.

The graph contains:

• 18 node types spanning investment entities (Investors, Fund Managers, Funds, Deals, Portfolio Companies), intent signals (Appetites, Live Intentions, RFPs, Commitments), controlled vocabularies (Strategy, Sector, Region, Topic), editorial content (Articles, Podcasts, Events), and people

• 85+ nodes including 12 investor organizations, 8 fund managers, 15 funds, 22 deals, 18 portfolio companies, and representative samples of intent signals and editorial content

• 200+ relationships that connect investment chains, map intent signals to strategies/sectors/regions, link people to organizations, and bridge editorial content to data entities

• Controlled vocabularies with 8 investment strategies, 8 sectors, 6 geographic regions, and 12 editorial topics—enforced through the graph structure to ensure consistency across all products

Key modeling decisions:

• Investment strategy, sector, and region are inferred from actual deal behavior, not just declared fund focus—enabling "strategy drift" analysis

• Intent signals (appetites, intentions, RFPs) are modeled as discrete nodes with lifecycle metadata, not just current state

• Articles connect to the same controlled vocabularies as business entities, solving PEI's core challenge: bridging editorial taxonomy to data classification

This isn't a proof-of-concept with placeholder data. Every node has realistic properties, every relationship reflects actual private equity industry patterns, and queries return meaningful results that demonstrate cross-product intelligence.