Master Power BI from first principles to Fabric & Copilot AI.

At its core, Power BI does four things really well: it connects to almost any data source, transforms messy data into clean tables, models those tables into a relational structure you can query with formulas, and visualizes the results as interactive reports and dashboards that can be shared securely across an organization.

Why organizations use it

The three faces of Power BI

Core capabilities at a glance

• Data connectivity: 200+ native connectors — Excel, SQL Server, SharePoint, Azure services, Salesforce, Google Analytics, web APIs, and more.
• Transformation & modeling: Power Query for shaping data; the Tabular model + relationships for structure.
• Interactive visuals: Bar/line/map/matrix visuals plus a marketplace of custom visuals (AppSource).
• Natural language: Q&A and Copilot let you type or speak a question and get a chart back.
• Embedded & automation: Embed reports in apps with Power BI Embedded; automate refreshes and alerts.
• Governance: Row-level security (RLS), object-level security (OLS), sensitivity labels, audit logs, deployment pipelines.

The end-to-end data flow

1. Connect

   Power BI Desktop connects to one or more sources (databases, files, APIs, dataflows) using a connector and a connectivity mode: Import, DirectQuery, Live Connection, or the newer Direct Lake.

2. Transform

   Power Query Editor applies a recorded sequence of transformation steps (the "M" language under the hood) — filtering rows, fixing types, merging tables — before the data lands in the model.

3. Model

   Tables are connected through relationships into a star schema: fact tables (transactions) surrounded by dimension tables (date, product, customer). DAX measures are written against this model.

4. Visualize

   Fields and measures are dragged onto a report canvas, producing charts, tables, maps, and KPI cards that respond to filters, slicers, and cross-highlighting.

5. Publish & share

   The .pbix file is published to a workspace in the Power BI Service, where it can be turned into an app, embedded, scheduled to refresh, and shared with the right audience under the right permissions.

6. Consume & act

   Viewers explore dashboards on web or mobile, set data alerts, subscribe to email snapshots, or ask Copilot a follow-up question in plain English.

Connectivity modes compared

Where Power BI fits in the Microsoft data stack

Categories of sources

Key skills to master

• Choosing Import vs. DirectQuery based on data volume, freshness needs, and DAX complexity.
• Managing credentials per data source (Windows, Basic, OAuth2, Organizational account) and understanding privacy levels (Public/Organizational/Private) that control whether Power BI can fold queries across sources.
• Combining sources with Merge Queries (like a SQL join) and Append Queries (like a UNION).
• Parameters to make source paths or filters reusable and easy to change without editing every step.
• Gateways — the On-premises Data Gateway lets the cloud Service refresh data that lives behind a firewall.

Power Query Editor shows an Applied Steps pane on the right — every transformation (renaming a column, filtering rows, changing a type) is a small, replayable, and reorderable recipe. This makes your data prep transparent, auditable, and repeatable every time the data refreshes.

Everyday transformations

A taste of M

Every query is really just a chain of M function calls. A simple query that filters and adds a column looks like this:

let
    Source = Excel.Workbook(File.Contents("Sales.xlsx"), null, true),
    Sales_Table = Source{[Item="Sales",Kind="Table"]}[Data],
    FilteredRows = Table.SelectRows(Sales_Table, each [Amount] > 0),
    AddedMargin = Table.AddColumn(FilteredRows, "Margin", each [Revenue] - [Cost])
in
    AddedMargin

Performance: query folding

Query folding is what happens when Power Query can translate your applied steps back into a single native query (e.g. SQL) that runs on the source server instead of pulling all the raw data into memory first. Filtering and grouping early — before adding custom columns — keeps folding possible for as long as possible, which can turn a 10-minute refresh into a 10-second one.

Star schema, the gold standard

A star schema puts one or more central fact tables (things that happened — sales, clicks, tickets) at the middle, surrounded by dimension tables (descriptive context — Date, Product, Customer, Region). Facts are numeric and narrow; dimensions are descriptive and wide. This shape keeps relationships simple, DAX fast, and reports intuitive.

Relationships

• Cardinality: most relationships should be one-to-many (one row in a dimension relates to many rows in a fact).
• Cross-filter direction: Single direction is the default and usually the safest — Both directions can cause ambiguous filtering and slower performance if overused.
• Active vs. inactive relationships: only one active relationship can exist between two tables at a time; extra ones are marked inactive and activated inside DAX with USERELATIONSHIP() — handy for role-playing dimensions like Order Date vs. Ship Date.
• Avoid many-to-many where possible — it's supported, but adds complexity and performance cost; a well-designed bridge table is often a cleaner fix.

The date table

Almost every serious model needs a dedicated date dimension — a continuous calendar table marked as the official "Date Table" in Power BI. It powers time intelligence functions (year-to-date, prior period, rolling averages) and lets you build hierarchies like Year → Quarter → Month → Day.

DimDate = CALENDAR(DATE(2020,1,1), DATE(2026,12,31))

Snowflake vs. star

A snowflake schema normalizes dimensions further (e.g., splitting Product into Product → Category → Department across three tables). It saves storage but adds relationship hops that slow DAX and complicate reports. In Power BI, denormalizing back toward a flatter star schema is usually worth the extra storage.

Calculated columns vs. measures

Row context vs. filter context

Row context is DAX walking through a table one row at a time — it exists naturally in calculated columns and inside iterator functions like SUMX. Filter context is the set of filters currently applied by slicers, rows/columns in a visual, and the formula itself — it's what makes the same measure return a different number for every cell of a matrix.

CALCULATE — the most important function in DAX

CALCULATE() evaluates an expression in a modified filter context. Almost every advanced DAX pattern — year-to-date, percent of total, comparisons across a slicer selection — is built on top of it.

Total Sales YTD =
CALCULATE(
    [Total Sales],
    DATESYTD('Date'[Date])
)

Sales - Prior Year =
CALCULATE(
    [Total Sales],
    SAMEPERIODLASTYEAR('Date'[Date])
)

% of Category Total =
DIVIDE(
    [Total Sales],
    CALCULATE([Total Sales], ALL(Product[Category]))
)

Function families worth knowing

Common beginner mistakes

• Using SUM where a row-level calculation is actually needed (that's what SUMX is for).
• Dividing without DIVIDE(), which silently crashes visuals on a divide-by-zero — DIVIDE() handles it gracefully.
• Overusing calculated columns for things a measure could do dynamically, bloating model size.
• Forgetting that CALCULATE filters replace existing filters on that column by default, rather than adding to them.

Choosing the right visual

Design principles

Interactivity features

• Slicers & filters: page-level, report-level, and visual-level filters plus slicer visuals for on-canvas control.
• Cross-highlighting & cross-filtering: clicking one visual filters or highlights the others on the page.
• Drillthrough: right-click a data point to jump to a detail page filtered to that context.
• Bookmarks & buttons: capture a state of filters/visibility to build guided, app-like navigation.
• Tooltips pages: custom report pages shown as a rich hover tooltip.
• Conditional formatting: data bars, color scales, and icon sets driven by rules or measures.

Workspaces

A workspace is a container for related content — datasets (semantic models), reports, dashboards, and dataflows. Teams typically run separate Dev, Test, and Prod workspaces connected by a deployment pipeline, so changes can be validated before reaching end users.

Refresh & alerts

• Scheduled refresh — up to 8x/day on Pro, 48x/day on Premium capacity.
• Incremental refresh — only reprocesses new/changed partitions instead of the whole table, dramatically speeding up refreshes on large fact tables.
• Data-driven alerts — notify a user by email when a KPI crosses a threshold.
• Subscriptions — deliver a report snapshot to an inbox on a schedule.

Sharing content

Row-Level Security (RLS)

RLS restricts which rows a user can see in a dataset, based on their identity. You define roles in Desktop with a DAX filter expression, then map users or Azure AD groups to those roles in the Service.

-- Role: "RegionManager"
-- Table filter on Sales:
[Region] = USERPRINCIPALNAME()

-- Or, mapping via a security table:
[RegionID] IN
    CALCULATETABLE(
        VALUES(UserRegionMap[RegionID]),
        UserRegionMap[Email] = USERPRINCIPALNAME()
    )

Dynamic RLS (using USERPRINCIPALNAME() against a mapping table) scales far better than static RLS (a hard-coded role per region), since you manage access in a data table instead of republishing the .pbix for every change.

Object-Level Security (OLS)

OLS goes a level further than RLS: it hides entire tables or columns from a role, not just rows — useful for hiding sensitive columns like salary or cost from most viewers while still using them in a permitted measure.

Sensitivity labels & compliance

• Sensitivity labels (Confidential, Highly Confidential, etc.) travel with exported files (PDF, Excel, PowerPoint), enforcing protection outside Power BI itself.
• Microsoft Purview integration gives data lineage, cataloging, and audit trails across the whole estate.
• Certified & promoted datasets signal which semantic models are the trustworthy "single source of truth" versus a personal draft.
• Audit logs track who viewed, edited, or exported what, and when — essential for regulated industries.

What is Microsoft Fabric?

Fabric is a SaaS platform that unifies data engineering, data science, real-time analytics, and business intelligence on a single, shared foundation called OneLake — think of it as "one lake for the whole organization," with every workload reading and writing the same open Delta Parquet files instead of duplicating data between tools.

Direct Lake mode

Direct Lake is a new storage mode that lets Power BI read Delta tables in OneLake directly into the VertiPaq engine's memory format — without a separate Import step and without the query-per-visual cost of DirectQuery. It's the best of both worlds: near-Import speed with near-DirectQuery freshness, ideal for very large Fabric-hosted datasets.

Copilot in Power BI

• Report authoring: describe the report you want in plain English and Copilot drafts pages and visuals to start from.
• DAX query & measure assistance: ask for a calculation in natural language and get a working DAX starting point.
• Narrative summaries: auto-generate plain-English summaries of what a visual or page is showing, refreshed as data changes.
• Q&A evolution: Copilot builds on the older Q&A natural-language feature with much stronger context awareness.

Copilot works best on a clean, well-labeled, star-schema model with clear column and measure names — another reason good data modeling (Module 3) pays off everywhere else.

Featured tutorial

Beginner

• Power BI Full Course for BeginnersComprehensive multi-hour walkthrough from install to first dashboard
• Power BI Desktop Interface TourA guided tour of the Ribbon, Fields, and Visualizations panes
• Guy in a Cube (channel)Weekly tips straight from the Microsoft Power BI product team

Data modeling

• Introduction to Data Modeling in Power BIRelationships, cardinality, and star schema basics
• Star schema deep dives (search results)Curated search — pick the top result matching your source system

DAX

• DAX Fundamentals — SQLBIThe most respected deep-DAX teaching channel, by Marco Russo & Alberto Ferrari
• CALCULATE explainedMultiple creators break down filter-context modification with visuals

Fabric & Copilot

• Microsoft Fabric for Power BI usersBridging existing Power BI knowledge into the Fabric platform
• Copilot in Power BI — live demosSee natural-language report authoring in action

Official Microsoft samples

• Sample datasetsFinancials, Retail Analysis, Human Resources, Opportunity Analysis — free to download and rebuild
• Retail Analysis sample walkthroughSales & inventory analysis with drillthrough patterns
• Human Resources sample walkthroughHeadcount, attrition, and diversity dashboard patterns

Community galleries

• Power BI Data Stories GalleryHundreds of community-submitted reports with commentary from the authors
• Custom visuals SDK & samplesBuild or extend your own visuals beyond the out-of-the-box set
• Power BI Desktop sample filesDownloadable .pbix files straight from Microsoft's GitHub

What you'll find here

• Structured modules: a deliberate order from "what is Power BI" through to Fabric & Copilot.
• Practical explanations: real DAX, real M code, real scenario walkthroughs — not just definitions.
• Certification guidance: what each exam actually tests and how to prepare for it.
• Curated external resources: the best official docs, books, courses, and community links.
• Self-check quizzes: quick, honest feedback on what's actually sticking.

Find me elsewhere