<![CDATA[The exponential growth in data volume, velocity, and variety has exposed the limitations of traditional Extract–Transform–Load (ETL) approaches, which often struggle with scalability, latency, and computational inefficiency. This study aims to demonstrate the effectiveness of ETL optimization techniques by measuring ETL job processing time for a “fact credit card” dataset under various configurations: without optimization, with partitioning, with parallelism, and with a combination of both. This research adopts an experimental design. The study object consists of “fact credit card” transaction data, representing a large and complex dataset within a financial business environment. Data were collected through direct testing of different ETL process configurations. Data analysis was conducted using a comparative quantitative analysis approach. The results indicate that partitioning and parallelism each contribute significantly to reducing ETL processing time, with partitioning alone achieving a reduction of 79% and parallelism achieving a reduction of 20%. However, the combination of partitioning and parallelism delivers the highest performance improvement, reducing processing time by up to 91%. These findings underscore the importance of selecting and integrating appropriate optimization techniques to enhance ETL performance, particularly in large-scale data delivery environments such as financial services. Optimized ETL processes enable faster data availability, improved resource utilization, and enhanced scalability, all of which are critical for maintaining competitive advantage in data-intensive industries.]]>
