Overview

A financial services organisation with a rapidly scaled Power BI deployment engaged us to assess and optimise a reporting environment that was underperforming against expectations. We delivered a structured estate review, targeted remediation across a set of candidate reports, semantic model consolidation and a governance framework designed to protect the reporting investment long term.

The Challenge

Like many organisations that have adopted Power BI quickly at scale, the environment had grown faster than the practices supporting it. Reports were slow to interact with, semantic models were running up against service limits and end users had begun reverting to the legacy tooling they had been migrated away from.

Common post-adoption Power BI challenges we address in engagements like this include:

• Reports and models built without alignment to Microsoft best practices or Kimball dimensional modelling.
• Semantic models exceeding the 1 GB Power BI Pro model size limit, forcing unplanned licensing decisions.
• Inefficient DAX producing slow page loads and poor user interaction.
• One-to-one report-to-model patterns leading to a proliferation of redundant semantic models across the tenant.
• End users reverting to previously used tools because the new reports felt slower or less capable than expected.
• Absence of model endorsement, quality checks or lifecycle practices for newly created content.

Our Solution

We applied our proven Power BI Optimisation Review methodology, designed to stabilise scaled Power BI deployments and restore confidence at both the technical and end-user layer.

Discovery & Estate Assessment

We started with a full assessment of the tenant: workspace structure, semantic model inventory, report inventory, data source sprawl and usage telemetry. From that assessment we selected four candidate reports for deep-dive analysis - chosen to be representative of the estate's performance and modelling patterns so that remediation would transfer beyond the reports themselves.

Data Model Alignment: Star Schema & Kimball Methodology

Working with the client's data engineering team, we reshaped the underlying database views to align more closely with Kimball dimensional modelling. This produced conformed dimensions and cleanly defined fact tables, in place of the wide, reporting-shaped views the reports had been built against. A star schema is the recommended modelling pattern for Power BI because it allows the VertiPaq engine to compress efficiently and evaluate filter context predictably, which directly improves both report responsiveness and model size.

DAX Optimisation with DAX Studio

We extracted the DAX from the candidate reports, profiled measures using DAX Studio and VertiPaq Metircs and refactored the logic driving the slowest visuals. DAX Studio is the standard practitioner tool for diagnosing Power BI query performance: it surfaces storage engine versus formula engine time, expensive callbacks and inefficient iterator patterns. Measures were rewritten to reduce row context transitions, replace inefficient filter patterns and make better use of variables. On the worst-affected reports, page load times were reduced from over five minutes to under three seconds.

Semantic Model Consolidation & Size Reduction

The tenant contained more than ten redundant semantic models that were effectively duplicating the same underlying data at slightly different grains. We mapped the overlap, consolidated into a smaller set of shared semantic models, and repointed downstream reports. Alongside consolidation we reduced affected model sizes by more than 80% through column pruning, cardinality reduction, and removal of unused fields - bringing models that had previously exceeded the 1 GB Power BI Pro licence limit comfortably back within it and removing the need for blanket Premium Per User (PPU) licensing.

End-User Enablement & Adoption

Several end users had begun reverting to legacy tools because they believed Power BI was missing features they needed. In working sessions we demonstrated the Power BI capabilities that covered their use cases, including bookmarks, drillthrough, field parameters, personalised visuals and report-level filters. This closed the perceived feature gap and re-engaged users with the platform.

Governance, Quality Checks & Endorsement

To prevent recurrence of the same patterns, we put in place a lightweight governance framework: pre-publish quality checks covering model size, DAX patterns and relationship hygiene; a semantic model endorsement workflow so certified semantic models could be identified by downstream report authors and content lifecycle guidance covering workspace roles, app audiences and the XMLA endpoint for promoted models.

Technology, Features & Methodology Used

The engagement used Power BI Desktop and the Power BI Service alongside practitioner tooling including DAX Studio, VertiPaq Analyzer, Tabular Editor and Measure Killer. Data modelling work applied Kimball dimensional modelling, star schema design and conformed dimensions, with the reporting layer restructured around shared and certified semantic models, calculation groups and incremental refresh. Access and distribution were handled through row-level security (RLS), workspace roles, app audiences and the XMLA endpoint, with Power BI Pro, Premium Per User (PPU) and Microsoft Fabric (F-SKUs) licensing assessed as part of the model consolidation work.

Outcomes & Impact

• Delivered a full Power BI estate assessment covering workspaces, semantic models, reports and usage patterns.
• Aligned underlying data views to Kimball dimensional modelling, enabling consistent star schema patterns across the reporting layer.
• Optimised DAX across four candidate reports, reducing page load times on the worst-affected reports from over five minutes to under three seconds.
• Reduced affected semantic model sizes by more than 80%, bringing them back within Power BI Pro licence limits and removing the need for blanket PPU licence uplift.
• Consolidated more than ten redundant semantic models into a smaller set of shared, reusable models serving multiple downstream reports.
• Re-engaged end users who had reverted to legacy tools, restoring confidence in the Power BI platform and supporting wider adoption.
• Established governance, quality assurance and endorsement practices to protect the reporting estate as it continues to grow.