3. Model the Data (25–30%)
Phase 2 produced clean tables; this phase turns them into a semantic model that answers questions correctly and quickly. You'll configure how tables and columns behave (3.1), write the DAX measures that encode business logic (3.2), and find and fix what makes a model slow (3.3). This is where Phase 1's ideas about star schemas, filter flow, and evaluation context stop being theory: nearly every question in this domain can be solved by asking which way does the filter flow, and what context is this expression evaluated in?
Why it matters: a flawed model fails silently. Totals look plausible while double-counting rows or ignoring a slicer, and no error message warns you. Think of the model as plumbing: relationships are pipes that carry filters from dimensions to facts, and DAX decides which valves are open.
⚠️ Exam Trap: Bidirectional filtering and many-to-many relationships are rarely the best fix. The exam usually prefers a star schema with single-direction, one-to-many relationships.
