Abstract: Over the years, the size of data is increasing in organizations information systems. The value of information/data depends on its accuracy and timeliness. When it becomes obsolete or there is no need to use it, it becomes a burden on the system. In this study, our concern is improving the performance of data retrieval from relational database systems and to determine how long data should be retained on the system. We classified transaction data into two classes: basic data class and historical data class. We proposed new properties for data item and an algorithm for data management that help in data analysis and transferring data among system components automatically until it removed from the system. We implemented a prototype for the algorithm and measured data retrieval performance, we found that by reducing a table size the data retrieval performance is increased approximately by the same ratio.