4️⃣ Data Modeling (Critical for Interviews)

Data Modeling is the most important layer in Power BI after data cleaning. This is where you define how tables relate to each other, design the structure for optimal performance and analysis, and prepare the model for DAX calculations and report visuals.

A well-designed data model makes reports fast, scalable, and easy to maintain. A poorly designed model causes slow reports, incorrect numbers, circular dependency errors, and fails in interviews.

Power BI uses an in-memory columnar database called VertiPaq (part of the SSAS Tabular engine). Good modeling = better compression and faster query performance.

Macro: Relationships

Relationships define how tables are connected and how filters propagate across the model.

1. Creating Relationships

Methods:

• Auto-detect: Power BI sometimes creates them automatically (not reliable).
• Manual (Recommended):
  • Go to Model view (left sidebar).
  • Drag a column from one table to the matching column in another table.
• Manage Relationships dialog (Home → Manage Relationships) – best for overview and editing.

Cardinality and Cross-filter direction are automatically detected but should always be reviewed.

Key Requirements for a Relationship:

• Columns must have the same data type.
• One side should ideally be unique (no duplicates).

2. One-to-Many vs Many-to-Many

One-to-Many (1:*) – The Gold Standard

• Most common and recommended.
• Dimension Table (1 side) → Fact Table (* side).