A key step was migrating the transformation logic from the original Power BI models into the central data warehouse. This moved calculations and data transformations from individual reports into one managed environment. The result is one semantic model that serves as a shared foundation for further reporting and replaces the original 5–6 separate models.

The project was divided into several consecutive phases: setting up the Fabric environment, integrating data from source systems, building the data warehouse, creating the semantic model, setting up automatic refreshes twice a day, and preparing inputs for report migration and data validation. The client handled the validation independently, with support from the prepared outputs and data structure. 

A major challenge was that halfway through the project, the original Power BI solution stopped working. We therefore had to continuously release completed parts of the data warehouse and synchronize development between the DWH and Power BI reporting. As soon as a data section was ready, it was immediately published for validation and further report development. At the same time, the integration and transformation of other data areas continued in parallel. 

This approach allowed the client to start building new reports as early as possible, even on partially prepared data. The project therefore required fast decision-making, precise planning, and several technical compromises, especially in areas where it was necessary to prioritize data availability over ideal architecture in the first phase of the solution. 

The result is a centralized data warehouse in Microsoft Fabric that unifies data from key systems and provides a stable foundation for reporting in Power BI. The Power BI model refresh was reduced to approximately 1 minute, and thanks to the new architecture, data can be updated several times during the day. PKV gained a faster, more reliable and easier-to-manage reporting foundation for the further development of its analytics.