01Start with entities
Agree on what a customer, order, product, job, or return means.
Systems may represent the same real-world thing at different levels. One customer may have several emails, addresses, orders, support records, and accounting entries. The model must preserve those relationships without pretending every row is interchangeable.
- Source identifiers
- Cross-system matching rules
- Timestamps and reporting periods
- Status and category definitions
- One-to-many relationships
- Source lineage
02Normalize and validate
Consistency must be designed and tested.
Normalize names, dates, currencies, categories, and identifiers while retaining original source values. Reconcile totals and sample relationships against the systems people already trust.
If the unified model cannot explain a difference, it is not ready to become the business view.
03Report from the model
Build views around decisions, not source-system menus.
Once sales, returns, customers, products, projects, or campaigns relate coherently, reporting can answer cross-system questions without manually stitching exports together each time.