Power BI Fundamentals

• Name the parts of the Power BI stack and what each one owns
• Distinguish a measure from a calculated column and a report from a dashboard
• Trace a refresh from a source system all the way to a tile in the Service

• Excel answers one sheet at a time; Power BI answers across millions of rows with a single drag.
• A semantic model means analysts define the same numbers once, so every dashboard agrees.
• Governance lives in the Service: who can see which rows, and who can edit which datasets.
• It is the most widely deployed BI tool, so Power BI skills transfer across teams.

Power Query is the pantry prep (washing and chopping ingredients), the data model is the pantry and prep list (what is actually available and how items relate), DAX measures are the recipes (compute on demand), and reports are the plates served from the same kitchen — every plate relies on the same prepped ingredients.

• Power Query extracts, cleans and shapes raw data into tables you can model.
• The data model stores tables with relationships and compresses them for fast in-memory queries.
• DAX measures compute answers at query time — a measure is a recipe, not a stored number.
• The report canvas places visuals bound to measures and fields.
• Power BI Service stores the report, refreshes the model on schedule and enforces access rules.

• An ops team publishes one shared sales model; finance and field teams build separate reports that query the same measures, so revenue never disagrees.
• A European bank refreshes its regulatory cube every night at 5:00 UTC, with Service refresh logs paging an owner if a load slips past 6:30am.
• A BI team publishes a certified dataset with row-level security so regional managers only ever see their own territory's rows.

• Keep queries in Power Query and business logic in measures, never mixed randomly across tables.
• Name measures like units of analysis: 'Total Order Value (USD)' not 'Sum of amount'.
• Structure the model as a star schema: one fact table, small lookup dimensions.
• Publish one certified dataset and let reports connect to it instead of each report importing its own copy.

• Building labels and formatting inside code instead of in the model, so every measure duplicates logic.
• Keeping a giant flat table because 'joins are hard', then fighting duplicated row totals.
• Editing the published dataset in Service while fights happen in Desktop.
• Serving a dashboard from a laptop that must stay switched on 24/7.

• Check whether a number changes when you remove other visuals — that usually means a missing CALCULATE or a bad relationship.
• To see what Power Query produced, compare row counts at the 'Leave' vs 'Enter' step of a transformation.
• Use DAX Studio to see the query generated for a slow visual.
• Turn on performance analyzer in the View tab to time each visual.