Choose Import when refreshed data is current enough and the model fits your capacity; Microsoft recommends it as the default because it delivers in-memory query performance and broad modeling features. Choose DirectQuery only when data must remain at its source or a full import is impractical—and only after confirming that the source and connection can respond quickly under real report usage. Choose a live connection when a governed Power BI or Analysis Services semantic model already owns the measures, relationships, and security your report should use.
The key distinction: Import and DirectQuery describe how a semantic model accesses data, while a live connection describes a report consuming an existing semantic model. These choices affect freshness, performance, modeling flexibility, security, and operations.
What each Power BI connection mode does
Import
Import loads selected source data into the semantic model’s in-memory cache. Report visuals query that cached snapshot, which typically supports responsive interaction. Changes at the source appear only after the model refreshes, so the refresh schedule, data volume, capacity, and any gateway requirements must suit the report.
Microsoft’s current guidance is to use Import by default, then consider another approach when a concrete requirement makes it necessary.
#1 Best Overall
DirectQuery
DirectQuery leaves table data at the source. Visual interactions generate source queries, so displayed data can be closer to the source’s current state after a requery. It is not automatically real-time: visual refresh behavior and caching can affect what users see. Responsiveness also depends on the source, network, and any gateway in the path.
Microsoft identifies large data volumes, near-real-time requirements, and federated access as possible reasons to consider DirectQuery. It also brings query load to the source and has modeling and transformation limitations. Review Microsoft’s DirectQuery guidance before designing around it.
Live connection
A live connection lets a report consume an existing Power BI semantic model, Azure Analysis Services model, or SQL Server Analysis Services model. The report does not create its own local semantic model, so the upstream model remains responsible for its definitions and modeling. In Microsoft’s documented behavior, the user’s identity is passed to the upstream model for security trimming. Because the model is external, some modeling options are unavailable in the report.
Rank #2
Live connection is not another name for DirectQuery to a database. The remote semantic model may itself use Import or DirectQuery for its tables. See Microsoft’s comparison of live connections and DirectQuery.
Recommended Free Tools
Choose based on the requirement
| Requirement | Starting option | What to verify |
|---|---|---|
| Fast interaction and broad transformation flexibility; periodic refresh is acceptable | Import | Model size, refresh duration, capacity, and how old the data can be. Microsoft recommends Import by default. Microsoft Learn |
| The full dataset is too large or costly to import, or source queries are needed for current data | DirectQuery | Connector support, transformations, source response times, network and gateway overhead, concurrency, and security. Microsoft Learn |
| A governed enterprise model already defines measures, relationships, and access | Live connection | Permissions and whether the report’s modeling limitations are acceptable. Microsoft Learn |
| Recent facts need fresher values while older history can be cached | Hybrid table | Microsoft describes a DirectQuery partition for the latest data alongside imported historical partitions. Microsoft Learn |
| Large data is in a qualifying Fabric lakehouse or warehouse and low-latency reads are required | Direct Lake | Confirm the Fabric setup and applicable behavior; Microsoft’s decision guide identifies Direct Lake as an option for this case. Microsoft Learn |
| Repeated queries can use summaries, or remote latency and concurrency are bottlenecks | Import or aggregations over DirectQuery | Check that the model design and aggregation coverage match actual report queries. Microsoft Learn |
| Several external sources need to be combined without full ingestion | DirectQuery composite model | Assess source support, composite-model limits, and cross-source query behavior. Microsoft Learn |
These are starting points, not guarantees. Connector capabilities vary by source and mode; check the connector’s “Capabilities supported” section in Microsoft’s connector documentation.
Evaluate the trade-offs before committing
Freshness versus responsiveness
Ask how current report data must be. Import reflects the source at refresh time; DirectQuery can query the source on interaction, subject to requery behavior and caching. Then test the full report experience: a single fast query does not establish how multiple visuals, filters, or concurrent users will perform.
Scale and refresh feasibility
For Import, determine whether the model can be loaded and refreshed within your capacity and operational constraints. A large row count alone does not prove that DirectQuery is the better choice: source query cost, report patterns, and refresh feasibility all matter.
Modeling flexibility
Import generally offers the broadest feature set and transformation flexibility. DirectQuery has restrictions that can affect modeling and transformations, so verify that the required design is supported before selecting it. A live-connected report uses the upstream model rather than defining an independent local model.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Security and operational ownership
Decide which layer should own credentials, source permissions, and access rules. For a live connection, confirm permissions on the upstream semantic model and how identity is handled. For DirectQuery, validate source security and expected query load. For any published model, check whether credentials and an on-premises data gateway are needed.
Rank #4
Connectivity and usage conditions
Confirm the connector supports the selected mode, and test with representative visuals, filters, row-level security, gateway paths, and expected concurrency. A design that works for one author on a local connection may behave differently when published and used by a team.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.DirectQuery performance limits to know
Microsoft’s Power BI Desktop DirectQuery guidance recommends source responses of five seconds or less for visual data. It describes responses taking more than 30 seconds as producing an unacceptably poor experience, and says a query longer than four minutes times out in the Power BI service. These are Microsoft recommendations and service behavior—not performance measurements for a particular source or model. See the DirectQuery performance guidance.
Use these figures as checks when testing, not as a substitute for testing your own report. Measure representative visual queries and filters, and include the network or gateway and realistic user load in the test.
Free tools Windows power users keep installed
One-click scans. No signup required.
Best Value
Consider alternatives to pure DirectQuery
- Hybrid tables: Keep recent data in a DirectQuery partition while storing historical data in imported partitions.
- Direct Lake: Consider it for large data in a qualifying Fabric lakehouse or warehouse when low-latency reads are needed; verify the relevant Fabric configuration.
- Aggregations: Use imported summaries to answer common queries while retaining DirectQuery for detail where the model design supports it.
- Composite models: Combine data or connection approaches when a report needs to federate sources, while checking cross-source behavior and limits.
Microsoft’s decision guidance discusses these options as ways to address particular workload needs; they still require validation against the model and reports you plan to publish.
Check reversibility before changing storage modes
Do not assume a storage-mode change is easy to undo. Microsoft documents that, in Power BI Desktop, a table changed from DirectQuery to Import generally cannot be switched back to DirectQuery. The documentation notes specific exceptions for version-control scenarios in web modeling and live editing. Check the current storage-mode documentation before making a change that could constrain later design choices.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
