© Bmatix 2025 www.bmatix.be 1 Implementing Microsoft Fabric at Joël Vangrunderbeeck Eager to learn, happy to share... BI Consultant @Bmatix Find me on LinkedIn! Gorik Vandersanden Empowering innovation , driving progress Founder Sparki BV Find me on LinkedIn! © Bmatix 2025 www.bmatix.be 2 1. Intro Sparki 2. Previous Reporting Approach 3. Reporting Needs 4. Why Microsoft Fabric 5. Fabric Architecture 6. Lessons Learned & Struggles 1 2 3 4 5 6 © Bmatix 2025 www.bmatix.be 3 Intro Sparki Our Mission: The most powerful network in Belgium Sparki will become the largest network of Ultra - fast chargers in Belgium with 1,200 charging points by 2026 Electric charging at the intersection of speed and retail Sparki invests exclusively in Ultra - fast charging (320 - 720kW) Customers can charge up to 80% in 15 to 30 minutes Ultra - fast charging close to you on locations with additional activities and with a national coverage Where are we standing today ? 75 locations operational – 155 UFC’s – 312 UF charging points 900 charging session per day – 27,000 kWh sold per day 300 locations signed for installation 1 2 3 4 5 6 Intro Sparki © Bmatix 2025 www.bmatix.be 4 Previous Reporting Approach Data Engineer collecting data via C# scripts in a PostgreSQL Local to Cloud move No way of following costs Multiple reporting tools Reporting resides inside specific tools Different logins Data is scattered Not reporting multiple sources together Different tools to manage Only PROD environment 1 2 3 4 5 6 Previous Approach © Bmatix 2025 www.bmatix.be 5 Reporting Needs Sources Cloud Data Sources • DeftPower – CSP CDR ’ s • Greenflux – MSP CDR ’ s • ZOHO – Project Planning tool • Exact Online • Dynamics • ... 1 2 3 4 5 6 Reporting Needs © Bmatix 2025 www.bmatix.be 6 Reports • Combining data from different sources • Overview of charging stations: cost & revenue • Tool specific reporting like helpdesk 1 2 3 4 5 6 Reporting Needs © Bmatix 2025 www.bmatix.be 7 Platform Requirements • Simple to start • Scalable • Extensible – Data science in future • Evolutive / up to date with newest technologies • SAAS 1 2 3 4 5 6 Reporting Needs © Bmatix 2025 www.bmatix.be 8 Why Microsoft Fabric ? A vast landscape... 1 2 3 4 5 6 Why Fabric ? © Bmatix 2025 www.bmatix.be 9 Why Microsoft Fabric ? 1 2 3 4 5 6 Why Fabric ? Easily Scalable & Cost Efficient Unified Data Platform (SAAS product) • Simple • Seamless integration with Microsoft suite (PAAS) • Built in Security & Governace Extensible Data Integration © Bmatix 2025 www.bmatix.be 10 Easily Scalable & Cost Efficient • Cost efficient and clear! • One License • Easily scalable (up) • Easy to calculate ! 1 2 3 4 5 6 Why Fabric ? Microsoft Fabric Capacity Calculator (preview) © Bmatix 2025 www.bmatix.be 11 Unified Data Platform & Extensible Data Factory Data Engineering Data Warehouse Data Science Real - Time Intelligence Power BI Partner & Industry workloads Copilot in Fabric OneLake Microsoft Purview 1 2 3 4 5 6 Why Fabric ? © Bmatix 2025 www.bmatix.be 12 Data integration • Fabric link for Dynamics • Other connectors out of the box • On prem & Cloud 1 2 3 4 5 6 Why Fabric ? © Bmatix 2025 www.bmatix.be 13 Fabric Architecture @ Sparki • Lakehouse set up with multi - layered Medallion structure • One data workspace , multiple reporting workspaces • OneLake Notebooks & Pipelines, Semantic model with Direct Lake Connection • SQL Analytics endpoint • Deployement pipelines & GIT 1 2 3 4 5 6 Architecture © Bmatix 2025 www.bmatix.be 14 Extraction Layer • Fabric link for dynamics • Notebooks + pipelines • Use parameter files • Data saved in files instead of tables! 1 2 3 4 5 6 Architecture © Bmatix 2025 www.bmatix.be 15 Data Quality Layer • Data quality scripts creating tables from files with up to date data • Dates, Regex, ... • Should contain the same data as source systems 1 2 3 4 5 6 Architecture © Bmatix 2025 www.bmatix.be 16 Transformation Layer • Transformations needed • Joining tables • Aggregating where needed 1 2 3 4 5 6 Architecture © Bmatix 2025 www.bmatix.be 17 Model Layer • Creating Fact and Dimension tables • Joining tables from different source systems • Ready for reporting • Understandable for end users 1 2 3 4 5 6 Architecture © Bmatix 2025 www.bmatix.be 18 Presentation Layer 1 2 3 4 5 6 • Power BI semantic models using Direct Lake connection • Up to date data • No worrying about Semantic model refreshes • All transformations are done • Easy reporting • Ad hoc analysis using ... • SQL analytics endpoint • Excel • Other tools if needed Architecture © Bmatix 2025 www.bmatix.be 19 Lessons learned • Master data is key – Think before you do • Sources change – and are not always as documented as they should • Always check source systems if you find anomalies • Pyspark where posible, SparkSQL if easier/faster • DataFlow Gen 2 ’ s consume more capacity – a lot more... • Naming convention – Choose something and stick by it • Develop for reporting not for LinkedIn influencers /Microsoft 1 2 3 4 5 6 Lessons & Struggles © Bmatix 2025 www.bmatix.be 20 Feel free to contact us joel.vangrunderbeeck@bmatix.be gorik@sparki.be LinkedIn