“Miner”, “Manager”, and “Researcher”: Three modes of analysis of scanner data

Abstract

In this paper we describe three stylized representations of how analysts approach the task of analyzing scanner data. We call these three modes of analysis the “Miner”, the “Manager”, and the “Researcher”. The “Miner” does a broad-based search for all interesting events in the market. He searches for the causes of the events using rank-ordered hypotheses about what causes market share to change. The “Manager” does a more narrow, but deeper analysis. He knows a priori that an event is of interest to management. The “Researcher” analyzes scanner data in an opportunistic fashion, using the data as a research tool.

We also describe the Phase 1 version of Scan?expert, an expert system based on the “Miner” mode of analysis. Scan?expert integrates knowledge about the effects of marketing actions with a measure derived from information theory in order to “notice” interesting events in a brand's marketing environment. In addition to finding events of interest to managers, Scan?expert uses a hypothesis-driven analysis strategy to find key factors in the data associated with interesting events. The expert system also uses rules to derive key competitors in different markets, and explains results of analyses in marketing terms. Scan?expert is still an evolving system. Future implementations will be developed to meet users' needs and to take advantage of additional information becoming available in scanner databases.