- 29 Aug 2025
- Kevarnold
- Comments: 0
In the first post of this series, we explored how interactive reports provide a safe and scalable entry point into self-service BI. Now, we take the next step: enabling users to build their own reports using semantic models.
This level of self-service offers more flexibility and insight—but it also requires more structure, training, and governance.
🧱 What Is a Semantic Model?
A semantic model is a curated, governed dataset that defines business logic, relationships, and calculations in a way that’s easy for users to understand. Think of it as a trusted foundation for building reports—without needing to write complex queries or understand the raw data.
In Power BI, these models are often published as certified or promoted datasets, which users can connect to directly in the Power BI Service.
🧰 Why This Level Matters
Building reports from a semantic model empowers users to:
- Explore data beyond what’s available in pre-built reports
- Create custom views tailored to their needs
- Answer ad hoc questions without waiting on a central BI team
It’s a powerful step toward data democratization—but it must be done responsibly.
🎓 Training, Documentation, and a Phased Approach
As discussed in the SML meeting, many users leave introductory training like “Dashboard in a Day” excited but unprepared for deeper self-service. That’s why organizations should offer:
- Role-based training tailored to different user types
- Hands-on workshops focused on building from certified datasets
- Guided practice in the Power BI Service to reduce complexity
But training alone isn’t enough—documentation is critical. To help users build confidently and correctly:
- Document the semantic model structure and logic
- Expose definitions for visible columns so users understand what each field represents
- Publish the DAX behind each measure, ideally with plain-language explanations
This transparency builds trust in the data and helps users avoid misinterpretation. It also reduces support overhead by answering common questions up front.
Encouraging users to start building in the Power BI Service makes it easier to get started and keeps the experience streamlined. As users grow more confident, they may transition to Power BI Desktop for more advanced capabilities. At that point, it’s important to provide additional training and clear guidelines.
For example, when users create report-level measures in Desktop, they should follow naming conventions—such as prefixing them with an underscore (e.g., _Custom Margin)—to ensure those measures can later be migrated into the semantic model without breaking existing reports.
This phased approach supports growth while maintaining consistency and governance.
🛡️ Governance in Action
At this level, governance becomes even more critical. Key practices include:
- Using certified/promoted datasets to guide users to trusted sources
- Implementing a workspace request system to manage publishing
- Disabling personal workspace publishing to avoid data silos
A Center of Excellence (CoE) can help define policies, while a Community of Practice (CoP) supports peer learning and shared standards.
🤝 Support for Power Users
Not every user will want to build reports—but some will thrive in this space. These emerging power users are valuable allies. Support them with:
- Office hours and 1:1 coaching
- Recognition and showcases of their work
- Opportunities to contribute to the CoP
By nurturing this group, you create internal champions who can help scale self-service responsibly.
💬 Clear Communication Is Key
It’s important to set expectations:
- Not all users will—or should—build reports
- Self-service doesn’t mean “no support” or “no rules”
- The goal is to empower users within a governed framework
💡 Final Thoughts
Building reports from a semantic model is where self-service BI starts to scale. It gives users the freedom to explore and create—without compromising trust or consistency. With the right training, governance, and support, this level of self-service can unlock tremendous value across the organization.
In the final post of this series, we’ll explore the most advanced level: enabling users to build their own semantic models.
This post was based on a meeting transcript and developed into a blog series with help from Microsoft Copilot.