Abstract
As organizations increasingly rely on data-driven decision making, the design of business intelligence (BI) systems has become a widely debated topic within computer science and information systems. In particular, there is ongoing discussion regarding how data warehouse modeling choices and ETL design influence the performance, reliability, and usability of analytical platforms, especially in cloud-based environments. While prior research has explored dimensional modeling, ETL processes, cloud databases, and BI tools independently, fewer studies examine how these components interact within a single, end-to-end analytics pipeline.
This project addresses this gap by evaluating the impact of dimensional data warehouse design on analytical query performance, data reliability, and dashboard usability in a cloud-hosted BI system. An end-to-end pipeline was developed using Python for ETL processing, PostgreSQL as a cloud-based data warehouse, and Power BI as the visualization layer. Two database schemas were implemented: a dimensional star schema and a non-dimensionalized relational schema. Both schemas were populated with the same dataset to enable direct comparison.
The results demonstrate that dimensional modeling improves query efficiency for aggregation heavy workloads and supports more responsive and interpretable dashboards. ETL validation and the use of Slowly Changing Dimensions further enhanced data reliability without introducing significant performance overhead. This study argues that data engineering decisions made prior to visualization play a critical role in BI system effectiveness and that dimensional modeling remains a practical and scalable foundation for modern analytics platforms deployed in the cloud.
© 2026 Kaylee Bautista, published by University of North Georgia
This work is licensed under the Creative Commons Attribution 4.0 License.