Power BI has become part of Fabric, and it brought a new data connectivity mode with it: Direct Lake. That mode removes a trade-off people have lived with for years.
01. The Old Dilemma
Import mode is fast but freshness depends on the refresh schedule. DirectQuery is fresh but goes to the source on every interaction and is slow. On large data sets both are uncomfortable.
02. What Direct Lake Does
It reads Delta tables in OneLake directly at in-memory speed. There's no copy and no refresh step, yet performance is close to Import mode. Data appears in the report the moment it's written to the lake.
03. Its Conditions
The data must sit in a Fabric lakehouse or warehouse table. You can't connect to external sources with Direct Lake. Some modelling features aren't supported and the model can fall back to DirectQuery.
04. When Import Should Stay
For small, rarely changing data sets, Import is still the simplest answer. There's no reason to move existing reports that do complex transformation in Power Query.
05. Model Design Doesn't Change
Whatever the mode, you still need a star schema, correct relationships and clean measures. Direct Lake does not speed up a badly built data model.
06. When Migrating
Move a single report to Direct Lake first and measure performance and freshness. Decide on a bulk migration only after that measurement.