Kevin Arnold

In the first two posts of this series, we explored how organizations can empower users to consume and build reports using governed datasets. Now we turn to the most advanced—and potentially transformative—level of self-service BI: enabling users to build their own semantic models.

This level offers maximum flexibility and ownership, but it also introduces the greatest need for governance, training, and cultural alignment.

🧠 What Does It Mean to Build a Model?

Building a semantic model means defining the data structure, relationships, calculations, and business logic that power reports. This includes:

• Connecting to source systems
• Shaping and transforming data
• Creating measures and calculated columns
• Defining hierarchies and perspectives

In Power BI, this typically happens in Power BI Desktop and is published to the Service for broader use.

🧰 Who Should Be Building Models?

Not everyone. As discussed in the SML meeting, only a subset of users will have the skills, interest, and business context to build models effectively. These are often analysts, data-savvy business users, or members of a decentralized BI team.

Organizations should:

• Identify and support these advanced users
• Provide clear criteria for who can publish models
• Offer pathways for collaboration with central BI or data engineering teams

🧱 Model Management Best Practices

To ensure consistency and maintainability, consider the following practices:

• Create simplified models or perspectives tailored to different user groups
• Use deployment pipelines to manage changes and promote models across environments
• Leverage tools like Semantic Link Labs and Tabular Editor to streamline model development, documentation, and deployment
• Establish naming conventions and folder structures for measures and tables

These practices help reduce duplication, improve discoverability, and support long-term scalability.

📚 Documentation and Definitions

At this level, data definitions and trust become even more critical. Users must be aligned on what each metric means and how it’s calculated. To support this:

• Maintain a business glossary and data dictionary
• Document DAX logic and model assumptions
• Use the Center of Excellence (CoE) to resolve disputes and define authoritative metrics

Transparency builds trust—and trust is essential when users are building the foundation others will rely on.

🧑‍🤝‍🧑 Culture and Communication

Building models is not just a technical task—it’s a cultural shift. Organizations must:

• Clearly communicate what “self-service” means at this level
• Acknowledge that mistakes will happen, and treat them as learning opportunities
• Encourage open dialogue between business and technical teams

Some users will prefer to delegate model building, and that’s okay. The goal is to support those who are ready while maintaining a strong governance framework.

💡 Final Thoughts

Self-service model building is the pinnacle of BI empowerment. It enables domain experts to shape the data they know best—but it also requires maturity, discipline, and collaboration.

By investing in training, governance, and cultural alignment, organizations can unlock the full potential of their data—and their people.

---

This post was based on a meeting transcript and turned into a blog series with help from Microsoft Copilot.