Data flows through three stages:

Sources → Intermediate → Core

Sources normalize raw data from each integration through the base → normalized → intermediate → source union pattern. Each source is self-contained and follows the same structure. See Sources.

Intermediate tables are cross-source unions and concept-specific processing steps, organized by concept:

• intermediate/entities/ — identifier and trait EAV unions
• intermediate/relationships/ — relationship declaration unions
• intermediate/measurements/ — measurement EAV unions
• intermediate/identity-resolution/ — entity matching and deduplication
• intermediate/states/ — state processing
• intermediate/attribution/ — touchpoint and path assembly

These are EAV canonical forms and processing artifacts — not consumer-facing.

Core tables are the output. They are the wide, pivoted, resolved tables shown in the core tables ERD. This is what analysts, LLMs, and downstream tools query.