How a Financial Services Organization Reduced Power BI Capacity Utilization by 35% Without Increasing Infrastructure Spend?
Semantic Model Redesign | Refresh Optimization | Capacity Management
At a Glance
| Industry | Financial Services |
| Services | Power BI Consulting, Semantic Model Optimization, BI Platform Management |
| Challenge | Rising Power BI infrastructure costs and increasing capacity utilization as adoption expanded across the business |
| Solution | Semantic model consolidation, refresh schedule optimization, and implementation of a capacity governance framework |
| Key Result | 35% reduction in peak capacity utilization, stabilized infrastructure costs, and continued platform growth without additional capacity investment |
About the Client
The client in question is a rapidly expanding financial services company that has been able to leverage Power BI effectively to develop its capabilities in departmental reporting into an enterprise-grade analytics solution.
Over the course of a year and a half, the use of Power BI extended into the Finance, Risk, Operations, Compliance, and Commercial departments. Although the implementation generated considerable value for the company, increased usage also resulted in greater costs incurred due to high infrastructure costs and poor performance.
The cost of Power BI premium capacity had gone up by more than 40% in a single year, while daily usage levels were consistently exceeding thresholds that had a negative impact on reports' performance. The organization needed a way to control costs, improve performance, and support future growth without investing in additional capacity.
The Challenge
As the platform expanded, several issues emerged:
Inefficient Semantic Models
Over thirty datasets had already been created, some developed separately by different teams. It meant duplication of data, overlap of tables, and additional calculation columns that consumed extra memory space and refresh cycles while providing no added value from the business perspective.
Uncoordinated Refresh Schedules
Refresh schedules for the datasets had been set up separately by different report owners without much central coordination. This led to 18 datasets being refreshed between 7:00 AM and 9:00 AM—exactly when all other reports were run by the end-users who demanded a lot of resources for their tasks.
Lack of Cost Visibility
It was impossible to define where the capacity resources went because there was no way to track which departments, reports, or datasets used them. It made it hard to reduce costs or even discuss any issues related to them.
Dataset Proliferation
Teams did not try to reuse existing datasets and preferred to create new ones for their purposes. This led to multiple datasets containing similar data, each with its own refresh schedule, data pipeline, and maintenance overhead.
Our Approach
Aegis Softtech conducted a comprehensive review of the Power BI environment and implemented a structured optimization program focused on three areas:
- Semantic model optimization and consolidation
- Refresh process redesign
- Capacity governance and monitoring
Phase 1: Capacity and Model Assessment
First, we undertook an in-depth assessment of Power BI, which included:
- An analysis of 34 datasets in use and their behavior based on Power BI activity logs
- Assessment of refreshes against capacity utilization
- Memory profiling of the largest datasets via VertiPaq Analyzer
- Development of a cost attribution model to understand capacity consumption by department
Phase 2: Semantic Model Consolidation
To eliminate redundancy and increase efficiency, the semantic model architecture has been redesigned.
Key initiatives included:- Building four certified enterprise datasets for Finance, Risk, Operations, and Commercial disciplines
- Migrating department-specific datasets to reuse these certified models
- Replacing unnecessary calculated columns with measures where appropriate
- Eliminating redundant data storage and refresh activity
Phase 3: Refresh Optimization
We redesigned dataset refresh schedules based on actual business requirements rather than default refresh configurations. Improvements included:
- Distributing daily refreshes across a 14-hour window (5:00 AM to 7:00 PM)
- Removing the concentration of refresh activity during peak business hours
- Changing six datasets from daily to weekly refresh schedules where data freshness requirements allowed
- Implementing incremental refresh for large historical datasets, processing only new and changed records
Phase 4: Capacity Governance Framework
In order to achieve long-term sustainability, a governance model that emphasized accountability and transparency was adopted.
- Workspace Governance: New datasets now require review and approval before being deployed to Premium capacity, including an assessment of their expected capacity impact.
- Cost Attribution Dashboard: Executives had access to a dashboard that gave them insights into the consumption of capacity at a workspace, dataset, and departmental level.
- Capacity Monitoring: Automated alerts were configured at 70% and 85% utilization thresholds, allowing the team to proactively address potential performance issues.
- Dataset Ownership: Every dataset was assigned a business owner who was responsible for ongoing review, maintenance, and retirement when no longer needed.
The Results
| Metric | Before | After |
|---|---|---|
| Peak capacity utilization | Baseline | 35% lower |
| Morning peak refresh contention | 72% of the weekly load in 4 hours | Distributed across a 14-hour window |
| Active datasets in Premium workspace | 34 | 21 |
| Weekly refresh operations | Baseline | 42 fewer per week |
| Monthly capacity spend growth | +40% year-on-year | Stabilized |
| Capacity visibility | None | Real-time by workspace and department |
What Made the Difference?
Audit before buying more capacity.
While the natural inclination for the organization was to go for increased capacity, it was determined that the bottleneck stemmed from inefficiency, rather than from sheer volume, as the work could have been done at 35% reduced utilization without additional investments into infrastructure.
Fewer datasets mean less of everything.
Fewer datasets mean less of everything. Fewer datasets result in fewer pipelines, fewer refresh schedules, fewer people responsible for them, and fewer opportunities to fail. The benefit of the governance side of consolidation will pay off long after the savings have been realized.
Without governance, the problem comes back.
The conditions that caused the budget overrun in the first place would have returned within 12 months due to the lack of governance. It's the structure that ensures the lasting success of the optimization.
Technology Stack
- Power BI Premium Capacity (workload management, capacity metrics)
- Power BI Semantic Models / Datasets (consolidation and optimization)
- Power BI Activity Logs (usage analysis and cost attribution)
- VertiPaq Analyzer (model memory profiling)
- Incremental Refresh (large dataset optimization)
- Power BI Deployment Pipelines (governance and change management)
Our work doesn't stop here — we also build on .NET, .NET Core, Microsoft Fabric, and Azure Synapse to deliver complete data and application solutions.
Is Your Power BI Infrastructure Spending Growing Faster Than Your Business?
If capacity costs are increasing without a corresponding increase in analytical value, the cause is almost always model inefficiency, unmanaged refresh scheduling, or ungoverned dataset proliferation. Aegis provides structured capacity optimization and governance engagements.