Microsoft Fabric for Analysts: From Lakehouse to Direct Lake Model
Microsoft Fabric for Analysts is an advanced-level course of 5 lessons (about 1 h 40 min), with 15 exercises and 15 quiz questions. Prerequisites: Power BI Service: Publishing, Sharing and RLS Security. The course needs a Windows virtual machine with Power BI, planned for a later phase of the platform.
- Power BI
- Advanced
- 5 lessons
- 1 h 40 min
- 15 exercises
- Windows virtual machine (phase 2)
The course in brief
This course extends the Power BI path to Microsoft Fabric. You load data into a lakehouse and a warehouse, prepare it with Dataflows Gen2 orchestrated by a pipeline, build a gold layer in PySpark and Spark SQL, then create a Direct Lake semantic model without copying the data. You finish with governance: access, SQL security, labels, lineage, lifecycle and capacity.
What you will be able to do
- Explain OneLake and the Delta format, and choose between lakehouse and warehouse
- Prepare and load data with a Dataflow Gen2 and orchestrate with a pipeline
- Transform data and write Delta tables with a PySpark and Spark SQL notebook
- Create a Direct Lake semantic model and control how its data is updated
- Organize a medallion architecture from raw data to the report
- Govern a Fabric solution: roles, SQL security, labels, lineage, capacity
Syllabus: 5 lessons
- Place Fabric, OneLake and the items of a workspace in context
- Load files into a lakehouse and query them in SQL
- Choose between lakehouse and warehouse depending on the need
- 02
Dataflows Gen2 and Pipelines: Ingest and Prepare
- Prepare data with Power Query in a Dataflow Gen2
- Write the result to a destination (lakehouse or warehouse) with the right mode
- Orchestrate and schedule ingestion with a pipeline
- 03
Notebooks: Transforming with PySpark and Spark SQL
- Open a notebook attached to a lakehouse and read a table
- Transform and aggregate data with PySpark and Spark SQL
- Write Delta tables for the gold layer
- 04
Direct Lake: Semantic Models on OneLake
- Compare the Import, DirectQuery and Direct Lake storage modes
- Create a Direct Lake semantic model on the gold tables
- Understand how data is updated and the limits of Direct Lake mode
Direct Lake modeFabric semantic modelDelta table optimization
- 05
Governance: Access, Security, Lineage and Capacity
- Manage access through workspace roles, item sharing and SQL permissions
- Apply labels, endorsement and domains, and read the lineage
- Organize the lifecycle and monitor capacity consumption
Security in FabricGovernance and lineageCapacity and monitoring
Final project
Medallion Sales Analytics Platform
The retail chain from the Power BI path wants to move its reporting to Fabric. Build the complete chain: raw file in a lakehouse, silver layer through a Dataflow Gen2, gold layer (facts and dimensions) through a notebook, a Direct Lake semantic model with measures and RLS by region, a scheduled pipeline that chains everything and refreshes the model, and a governance note (access, SQL security, labels, lineage, capacity monitoring).
Deliverables
- The lh_ventes lakehouse with Files/brut, ventes_silver (12 rows) and the gold tables fait_ventes, dim_magasin, dim_produit
- The df_ventes_silver dataflow and the commented nb_gold notebook
- The scheduled pipeline dataflow → notebook → model refresh, with failure notification
- The ms_ventes Direct Lake model (Total ventes = 415; Lyon 148, Lille 135, Nantes 132) and a report
- A governance note: access matrix, GRANT/DENY statements, label, endorsement, lineage screenshot
Assessment criteria
- Replayable chain with no duplicates (Replace and overwrite modes justified)
- Exact results at every layer (12 rows, 415, breakdown by store and category)
- Clean Direct Lake model: relationships, measures, columns prepared upstream, controlled update
- Consistent security at all levels (workspace roles, SQL, RLS) and consumers without an editing role
- Governance and operations: lineage, endorsement, separate environments, attention to capacity
Frequently asked questions
Do you need to know how to code to use Fabric?
Not for the most part: lakehouses, Dataflows Gen2 and models are handled without code. Notebooks require Python or SQL basics, introduced gradually in the course.
What is the difference between Direct Lake and Import mode?
In Import, data is copied into the model and refreshed. In Direct Lake, the model reads the Delta tables in OneLake directly: no copy, and a refresh that only points to the latest version of the data.
Lakehouse or warehouse: which should you choose?
The lakehouse suits varied files and Spark teams; the warehouse suits SQL teams that need transactional writes. Both store their tables as Delta in OneLake and often coexist.
Which license do you need for Fabric?
It depends on your organization's license and capacity. Check with your administrator; the course does not detail the offerings, which change regularly.
This course, told for your job
Say what you want to be able to do: “Microsoft Fabric for Analysts” takes its place in your path, with examples from your field. The concepts stay those of the course.