We have implemented a high-capability system that gathers data as it processes user transactions. This data contains useful business information about those transactions, including who performed the transaction, when it took place, how it related to other transactions and how the system behaved during the transaction. By performing proper analysis on this data, the business can detect trends, for example, product sales and customer demographics. The operations team can also use such data to predict system load and detect potential overload. The system gathers information such as orders placed, duration of transactions, and page views. This information can be analyzed to determine which products were commonly bought together. This can be further refined when it is split on a demographic basis (for example, retailers who are identified as operating in a higher-income area). There may be many such analyses that are required--some standard and some ad hoc--as the business tries to determine the best combination of offers and marketing to suit the needs of its customers. Using live system data as the basis for such analyses puts an additional, intensive load on the system database and degrades the performance of the system