Microsoft Fabric changes the contract between your semantic model and the data underneath it. Direct Lake promises import-mode speed over lakehouse tables with no scheduled refresh — but only if you respect its rules. Get the modeling wrong and you silently fall back to DirectQuery, framing goes stale, and the “fast” model is anything but. This session is the practical guide to building Fabric semantic models that actually perform.

We’ll work through the decision that drives everything else — Direct Lake vs. Import vs. DirectQuery — and when each is the honest right answer. Then we’ll build a proper star schema over OneLake, cover the constraints Direct Lake imposes (data types, calculated columns, relationships), keep an eye on fallback, and apply the modeling and DAX patterns that hold up as data grows. You’ll leave knowing how to bring your Power BI modeling instincts into Fabric without getting burned by the new storage layer.