The problem
Your CRM has 320k company records but maybe 180k unique companies. Reps collide on accounts. Signals miss subsidiaries. Scoring writes to the wrong entity. Reports inflate pipeline.
The system
- Export to warehouse. Full CRM snapshot as the starting point.
- Resolve identity. Domain matching, name normalization, fuzzy matching with confidence scores.
- Build canonical table. One ID per company, all aliases mapped, subsidiaries linked.
- Sync back. Clean records in CRM, territory owners assigned, duplicates merged with activity preserved.
Field notes
This is the foundation project. Every other system (scoring, signals, routing, outbound) depends on a resolved universe.