Preliminary Investigations into Knowledge Discovery for Quick Market Intelligence

Quick Market Intelligence (QMI) is the process of identifying salient aspects of a business market to target sales more effectively. Retail businesses have begun to apply knowledge discovery techniques to their sales databases to quickly determine the optimal quantity of a certain product to stock, how well a sales promotion is doing, its location within a store, and its price, all given historical data obtained from previous sales. For example, a supermarket chain may stock certain food items only in certain stores, depending on the ~thnicity or age of the local population, or may restock a new item just-in-time given greater-than-anticipated sales volume. This paper reports on the results of applying existing software knowledge discovery methods (in particular, C4.5 and logistic regression) in support of a particular business application. The results, based on feedback from the GE business, indicate that this type of analysis is useful with very little special-purpose software development. We are currently using the results to better understand the tradeoffs between different discovery techniques, where extensions and research effort should best be spent,:and how the results can be transitioned to solve real business problems and improve on existing business processes.