As a long-standing leader in Business Intelligence, Power BI’s constant innovations are paving the way for other tools. The term ‘dataset’ in Power BI no longer fully describes its functionality. A dataset not only contains data but also transformations, calculations, and relationships as part of a complex and flexible model (semantic model). In this article, we will define semantic models, including what a semantic model represents and the various modes within it.

What are Semantic Models in Power BI?

A semantic model in Power BI can be viewed as a logical layer that contains transformations, calculations, and relationships between data sources necessary for creating reports and statistics. The semantic model serves as a single source of accurate information for reports across the organization.

Although a semantic model can be built using Power BI Desktop, it does not necessarily include visualizations. You can think of the semantic model as the final step in the data flow before reports and tables are created. When sharing the semantic model with other members of the organization, they can create multiple reports and statistics from that single semantic model.

Semantic models do not include the complex technical details behind reports, which allows users with different experience levels and expertise to focus on data analysis and answering business questions.

What Does a Semantic Model Contain?

Semantic models consist of several different elements:

• Connections to one or more data sources, either embedded via DirectQuery or as part of a composite model.
• Transformations that clean and prepare data for reporting.
• Defined calculations and metrics based on business rules to ensure consistency in reports created from the semantic model. This ensures clarity and avoids discrepancies between analyses and reports.
• Defined relationships between tables allow users to focus on report design.

Modes of Semantic Models

Choosing the right mode when connecting to data in Power BI is an important first step in creating a semantic model, as each mode has its own advantages and disadvantages. There are three modes of semantic models in Power BI:

1. Import Mode
2. DirectQuery Mode
3. Composite Mode

Import Mode

This mode fully loads data into the Power BI file. Each time a Power BI report is updated, the storage engine compresses, optimizes, and stores the data. This leads to fast performance and flexible design options for report creators. Additionally, Import Mode allows semantic model creators to use the full range of Power Query M functions for data transformation and preparation, as well as DAX functions for creating calculations and measures.

DirectQuery Mode

This mode stores only metadata about the model’s structure, not the data itself. When the model is queried (e.g., via visualization), data is retrieved from the underlying data source. This is especially useful for large data volumes or when there is a business requirement for near real-time data in reports.

Composite Mode

The Composite Mode represents a combination of Import and DirectQuery modes. This mode is useful when you need the power and performance of Import Mode along with the ability to view real-time data. A table can be set in Dual storage mode, allowing Power BI to choose the more efficient mode depending on the nature of the query.

ITC Consult has extensive experience in data analysis and visualization with Power BI across various business areas—such as analyzing sales performance, financial and production KPIs, and more. Our data analysis solutions start with a deep understanding of the goals and tasks for a specific business area. Our consultants are committed to this as a first step to understand your business needs and develop an appropriate solution.