Dynamic Population Segmentation in Online Market Monitoring

The objective of the SEMAMO (Semantic Market Monitoring) project is to make use of the increasingly growing information available at Web-based sales and marketing channels for market research, using semi-automatic analysis driven by application domain models. The assumptions are that (i) the Web may serve as a representative “picture” of reality, (ii) the respective online channels map salient market developments, and (iii) all of this accurately and in a timely manner. Limited server requests and market specific access structures of Web portals inhibit both full scans of sampling populations and random selection of sampled offers. Further, product feature categories entail multiple classifications within offer clusters (e.g., geography in tourism). Therefore, SEMAMO proposes an adaptive sampling strategy dealing simultaneously with (i) the dynamics of the population frame, (ii) price dynamics, and (iii) multiple (fuzzy) classifications of offered products. The paper discusses a heuristic method of dynamically segmenting monitored offer populations to stratify online data harvesting depending on both observed price changes and information relevance, and outlines the mechanics of harvest schedule derivation.