Data Modeling

Data models are trusted datasets — curated tables where all the business logic, joins, and transformations are already baked in. Business users see them as simple tables and don't need to understand the underlying complexity.

---

What is a data model?

A data model is a query that materializes into a table in your warehouse. It sits between raw data and the questions your team asks. Instead of every analyst re-deriving the same joins and filters, the logic lives in one place.

Two key properties:

1. Clear, consistent grain — every row represents the same thing (one order, one user-day, one transaction). No ambiguity about what a row means.
2. Fact / dimension form — models follow the classic pattern:
   • Fact tables — events and measurements (orders, page views, transactions)
   • Dimension tables — descriptive attributes (customers, products, regions)

---

Data models in MinusX

With MinusX, you can create data models directly from the UI. When you save a model, it is materialized into your warehouse / database as a real table.

Only data models can be whitelisted in a Knowledge Base for org-wide use. This ensures that all users across the organization query from the same trusted, governed datasets — not from raw tables with inconsistent logic.

---

Best practices

• Define grain first — before writing a model, decide what one row represents
• Keep models self-contained — a business user should be able to query the model without joining other tables
• Name clearly — fact_orders, dim_customers communicate intent immediately
• Document in context — add metric definitions and gotchas in the Knowledge Base so the AI writes correct queries

---