Microsoft Fabric: Build a Semantic Model

Key insights

• Semantic model in the video: Nick Lee shows how to build a first semantic model in Microsoft Fabric, connecting data, defining relationships, and preparing it for reporting, analytics, and AI-driven insights.
  He demonstrates the end-to-end web workflow so teams can start models without switching tools.
• Metadata layer explained: A semantic model organizes raw tables into clear business structures by adding measures, formats, and calculated columns.
  It uses entities and relationships so reports and visuals reflect business logic instead of raw tables.
• How to create a model: Go to a lakehouse or warehouse, select New Power BI semantic model, name the model and pick a workspace, then choose the tables (example: DimCustomer, FactOnlineSales) and generate the model.
  Open the model in the web editor to add relationships, measures, and display formats.
• Storage and query modes: Use Direct Lake to query Delta tables in OneLake without moving data, or choose Import or DirectQuery when needed.
  Combine types with Composite models for flexible performance and data freshness trade-offs.
• Integration and editing: Use web modeling to build and edit models in the browser, then create reports or connect via a live connection from Power BI Desktop so shared measures and relationships stay centralized.
• Benefits and best practices: Models improve scalability and consistency across reports, boost query performance, and reduce duplicate work by centralizing logic.
  Note: Fabric stopped auto-creating default models on September 5, 2025, so create models explicitly and favor Direct Lake for high-performance scenarios.

The video by Pragmatic Works, presented by Nick Lee, walks viewers through creating a first semantic model in Microsoft Fabric and explains why the model matters for reporting, analysis, and AI-driven insights. The tutorial emphasizes how a semantic layer organizes raw tables into business-friendly structures and connects to tools like Power BI so teams can build consistent reports. It also demonstrates the web-based workflow in the Fabric portal, showing how users can generate and edit models without immediately switching to desktop tools. As a result, the video frames semantic modeling as a foundational step for scalable analytics across an organization.

Getting started in the Fabric portal

The walkthrough begins with the practical steps to create a model from a data item such as a lakehouse or warehouse in the Fabric portal. Users choose the New semantic model option, name the model, pick a workspace, and select the tables they want to include; the interface then opens a web modeling view for immediate edits. Nick Lee shows how common tables like DimCustomer and FactOnlineSales can be added and how auto-detection or manual linking defines relationships between keys. Consequently, the portal flow reduces friction for teams who want to move quickly from raw data to a reusable analytical layer.

How the semantic model organizes data

The video explains that a semantic model acts as a metadata layer that maps physical data into business concepts, applying relationships, measures, and display formats. For example, Delta tables stored in OneLake or tables exposed via SQL analytics can be transformed into friendly table names, hierarchies, and calculated measures for reports. This abstraction helps analysts and report authors rely on a shared definition of metrics instead of recreating calculations across reports, which improves consistency. Moreover, the model supports different storage modes so teams can choose between live querying and importing data where appropriate.

Tradeoffs: performance, freshness, and complexity

One clear tradeoff the video highlights involves choosing between Direct Lake, import, and DirectQuery storage modes, each balancing performance and data freshness differently. For instance, Direct Lake can deliver high performance by querying Delta files directly without ingestion, but it may require careful tuning to meet concurrency demands; import mode can speed up queries at the cost of keeping data updated. Composite models offer flexibility by mixing modes, yet this approach can introduce complexity in debugging and governance. Therefore, teams must weigh immediate query speed against operational overhead, latency expectations, and governance needs when building models.

Challenges and recommended practices

Nick Lee also addresses common challenges like relationship design, naming conventions, and version control, all of which affect long-term maintainability of a semantic model. He recommends explicit relationship checks, consistent naming and formatting, and testing measures in both web modeling and Power BI Desktop to ensure behavior matches expectations. Collaboration practices matter too: centralizing the model in a workspace lets multiple authors reuse measures, but it requires governance policies to prevent accidental changes. In short, careful planning and clear processes reduce the risk of duplicated effort and inconsistent metrics across reports.

Implications for teams and governance

Finally, the video frames semantic models as a strategic enabler for enterprise analytics by enforcing shared definitions and improving report performance when done right. When teams adopt centralized models, they can scale reporting workflows and support AI-driven insights from consistent data structures, while also reducing the abundance of report-specific calculations. Yet success depends on training and role clarity, since web modeling is powerful but still demands thoughtful architecture and lifecycle management. As a result, organizations that pair Fabric’s web tools with governance and testing practices can deliver reliable analytics at scale.

In summary, the Pragmatic Works tutorial offers a clear, practical guide for creating a first semantic model in Microsoft Fabric, balancing how to move fast with how to keep models maintainable and performant. It shows the step-by-step portal workflow, explains the rationale behind storage choices, and highlights tradeoffs that teams should consider. Therefore, teams new to Fabric can follow the demo to build an initial model and adopt practices that help models grow into reliable, shared assets for reporting and analytics. Ultimately, the video positions semantic modeling as an essential foundation for scalable, consistent insights across an organization.

Keywords

Microsoft Fabric semantic model, Creating semantic model in Microsoft Fabric, Microsoft Fabric semantic model tutorial, Semantic modeling for Power BI in Fabric, Build semantic layer in Microsoft Fabric, First semantic model Microsoft Fabric guide, Fabric semantic layer best practices, Microsoft Fabric data modeling tutorial