Business insights with semantic models
1. Business insights with semantic models
We've seen how Fabric stores and queries data. Now let's look at how a semantic model turns that data into consistent, report-ready business insights.2. From raw data to insight
Raw data on its own is hard to use. The tables we saw earlier in this course, such as customers, products, and orders, hold related records, but nothing yet says how they fit together. A semantic model adds that meaning. It connects the tables and defines the business calculations. Reports then present the results to business users.3. What is a semantic model?
A semantic model organizes stored data for reporting. It connects related tables, defines standard business calculations, and provides the foundation that reports are built on. Where a notebook serves analysts who write code, a semantic model serves business users who need trustworthy numbers without writing any code.4. Relationships
A semantic model relies on relationships. A relationship connects two tables through a matching column. An Orders table might connect to a Products table through ProductID. That connection lets a report combine information from both tables. Without it, the tables remain isolated. With it, reports can answer questions such as sales by product category.5. Measures
A semantic model also defines measures. A measure is a business calculation defined once and reused everywhere, such as Total Revenue or Average Order Value. Because the calculation lives in the model rather than in individual reports, every report that uses it shows the same number. That consistency is what makes business users trust the results.6. Reports use semantic models
Reports are built on top of semantic models. The model defines how the data connects and how calculations work. The report turns that structure into charts, cards, tables, and filters. Business users can then explore the data without navigating the raw tables themselves.7. Direct Lake
To be useful, a semantic model needs data. The way a model gets its data is called its storage mode, and Direct Lake is one of those modes. Direct Lake reads the columns a report needs straight from OneLake, where the data already lives, without creating a separate imported copy. Queries are fast because those columns are held in memory, and there is no copy to refresh, so reports always reflect the current data.8. Alternative storage modes
A semantic model supports two other main storage modes. Import copies data into the semantic model for fast reporting. The data must be refreshed to reflect source changes. DirectQuery keeps data in the original source and queries it whenever users interact with a report. This provides fresher data but can be slower, because it relies on the source to do calculations.9. Copilot in analytics
That completes the model itself: tables connected by relationships, calculations defined once as measures, and reports built on top. Copilot is an AI assistant that works alongside these tools. It can summarize a report, suggest visuals, or draft content, making it easier to turn raw data into actionable business insights.10. Reviewing Copilot's work
Like any AI assistant, Copilot can make mistakes. That's why its output works best as a starting point. We review it and check the underlying data before relying on it. For high-risk, business-critical decisions, that human review is especially important.11. Let's practice!
Now let's practice working with a semantic model and the report built from it.Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.