Great reports start with great models—and great models are designed, optimised, and proven. In this full‑day, hands‑on workshop you’ll learn how to build Power BI semantic models that stay fast as they grow. We’ll begin with modelling fundamentals and quickly move into the strategies that separate good solutions from great ones: storage mode choices that scale, optimisation techniques that stick, and monitoring practices that keep you fast in production.

You’ll practise applying aggregations, hybrid tables, split columns, and data partitioning to tame large datasets. We’ll demystify distinct‑count at scale (both via data modelling and aggregation design), and you’ll blueprint models around common DAX patterns—period comparisons, ranking, and child‑to‑parent ratios—so they’re robust and repeatable. You’ll learn a pragmatic approach to diagnosing slow visuals and validating improvements with load testing using Microsoft Fabric (plus alternative options).

Finally, we’ll make your model Copilot‑ready with natural‑language semantics, synonyms, and well‑named artifacts; and you’ll operationalise performance with VertiPaq Analyzer, Log Analytics, Capacity Metrics, and BPA.
Along the way, you’ll work with the right tools for the job—Power BI Desktop, XMLA, TMDL/TMSL, and Semantic Link / Semantic Link Labs—to design, optimise, and prove your model end‑to‑end. You’ll leave with a tested performance playbook and reusable assets you can apply immediately.

Who should attend?
BI developers, data modellers, analytics engineers, and Power BI admins who already know basic DAX/modeling and want to ship faster reports, scale to bigger data, and prepare for Copilot.
What you’ll be able to do after this workshop
• Pick the right storage mode and partitioning strategy for your data size, refresh cadence, and SLA.
• Implement aggregations and hybrid tables that measurably reduce query time.
• Model distinct‑count accurately and efficiently, avoiding common pitfalls.
• Design for common DAX patterns (period comparisons, ranking, child‑to‑parent ratios) with predictable performance.
• Diagnose and fix slow visuals using a repeatable triage and tuning approach.
• Load‑test and monitor models using Fabric, VertiPaq Analyzer, Log Analytics, Capacity Metrics, and BPA.
Optimise for Copilot with natural‑language friendly semantics, synonyms, and artefact naming.