The ability to combine Direct Lake and Import in Power BI marks one of the most significant transformations in data modelling within Microsoft Fabric. This hybrid approach combines the performance efficiency of Import Mode with the scalability and near-real-time capabilities of Direct Lake, allowing enterprises to strike the optimal balance between speed, flexibility, and resource efficiency.

Let’s explore how web modelling supports mixed storage modes, the performance advantages it unlocks, practical use cases across industries, and best practices for implementation—both in Power BI Desktop and the Power BI Service.

Understanding Direct Lake and Import in Power BI

Traditionally, Power BI users selected a single storage mode—Import, DirectQuery, or Dual—per dataset. However, with the advent of Direct Lake, Power BI now allows seamless integration of Direct Lake and Import tables within the same semantic model, creating composite models that leverage the strengths of both modes.​

• Import Mode: Caches static data within Power BI for lightning-fast performance but requires scheduled refreshes.
• Direct Lake Mode: Reads data directly from OneLake Delta tables in near real time, eliminating heavy refresh operations and reducing redundancy.​
• Composite Models: Now allow modellers to import specific tables (for example, small lookups or historical data) while maintaining Direct Lake access for large or frequently updated datasets.​

Web Modelling with Mixed Storage Modes in Power BI

With the October 2025 Power BI update, you can now create composite semantic models using both Direct Lake and Import tables directly in the Power BI web experience. Web modelling introduces new UI elements to simplify these configurations, including:

• Four new web buttons for adding, converting, or replacing tables between storage modes
• The ability to switch specific dimensions or reference tables from Direct Lake to Import mode for features like calculated columns or Excel integration​
• Direct integration with OneLake, Fabric Lakehouses, Data Warehouses, and Power Query Online connectors

This mixed-mode modelling workflow enables teams to combine real-time analytics (Direct Lake) with stable, high-performance cached data (Import) within a single dataset—without needing multiple reports or datasets.

Performance Benefits and Use Cases

The mixed-mode approach offers measurable performance and architectural advantages for enterprises dealing with large data volumes, diverse sources, and complex relationships.

Key Benefits

1. Speed and Scale Without Refresh Overhead
   Direct Lake loads metadata instead of full datasets during refresh operations, drastically lowering bandwidth and compute costs compared to Import mode refreshes.​
2. Reduced Redundancy
   Import tables are used only when necessary (for example, small calculated dimensions), while Direct Lake eliminates full data duplication from OneLake.​
3. Real-Time Capabilities
   Business-critical tables can run in Direct Lake for live dashboards, combining freshness with interactive query speed similar to Import mode.
4. Improved Query Consistency
   Unlike classic DirectQuery-Import composites, Direct Lake + Import combinations maintain regular relationships between tables—ensuring predictable query performance.​

Implementation Best Practices

To fully exploit Direct Lake and Import combinations, Microsoft recommends the following design and optimisation practices:

1. Plan Data Partitioning Strategically

• Use Import mode for smaller, static, or dimension tables where transformations are required.
• Use Direct Lake for large fact tables that require near real-time updates from Fabric Lakehouses or Delta tables.
• Split historical and current data into separate tables—historical (Import), live (Direct Lake).​

2. Optimise Relationships

• Maintain standard relationships whenever possible to ensure smooth join propagation between mixed-mode tables.
• Configure Dual mode where flexibility between cached and live queries is required for complex relationships.​

3. Use OneLake as the Foundation

• Ensure all Direct Lake tables are stored in OneLake Delta format.
• Avoid redundant ETL pipelines by leveraging Fabric tools (Dataflows, Spark jobs, T-SQL DML) for pre-processing.​

4. Monitor and Tune Performance

• Use the Performance Analyser in Power BI Desktop to identify queries hitting Import or Direct Lake tables.
• Enable Aggregation Tables for hybrid models to accelerate recurring analytical queries on large Direct Lake datasets.​

5. Adopt Governance Controls

• Define data access roles in Fabric to control user permissions across mixed data storage layers.
• Implement version control for your semantic models using Tabular Editor or Fabric Git Integration.​

Power BI Desktop vs Service Capabilities

Capability	Power BI Desktop	Power BI Service
Model Creation	Supports live editing but limited table transformation for Direct Lake data	Full web modelling experience with conversion between Import and Direct Lake ​
Transformation	Power Query is limited to Import-only data sources	Power Query Online supports transformations for Import and composite models
Data Source Flexibility	Supports local file and Fabric lakehouse connections	Extends to SQL, Datamarts, and over 100 connectors ​
Performance Optimisation	Manual tuning via Performance Analyser	Automatic tuning and metadata-level refresh optimisation ​
Deployment	Publish-only	Direct modelling and management via Fabric workspace

Conclusion

The combination of Direct Lake and Import models represents a pivotal milestone in Power BI’s data architecture evolution. This capability empowers organisations to build high-performance, flexible, and scalable analytical solutions that adapt effortlessly to evolving data demands.

https://osmosys.co/contact-us/Ready to unlock the full potential of your data and digital transformation journey? Connect with the experts at Osmosys today!