Abstract:

This paper describes ANALYST, a knowledge-based system for analyzing panel data. ANALYST supports product managers and market researchers in analyzing market data, helping them in the primary problems of marketing decision making: detecting significant deviations of product sales from long-term expectancies or trends. If deviations are found, the system gives hints for explaining them using rules of thumb. As a user oriented, intelligent interface to a common database, it provides an overview of data concerning markets and products over time. The core of the system is a market model, derived from year long experience of experts in marketing and market research. Defining market models mathematically is difficult and in many cases computationally untractable (see Lilien & Kottler, 1983). Therefore, the market is modeled heuristically, using expert's rules of thumb. The result of analyzing verbal protocols of about 40 hours of interviews is a dependency network (product model), which covers connections between the main facts that describe the behavior of a product - and actions concerning the product - in the market. ANALYST's architecture is highly modular. The system is programmed in Smalltalk/V and runs under WINDOWS on PCs. ANALYST was developed as a joint research project between the information management section, subsection market research, of a large European food company and the Department of Applied Psychology of the University of Regensburg, Germany.