Papers: Designing geodemographic classifications to meet contemporary business needs

Abstract

With statistics from the UK 2001 Census now available for small areas, the many marketers who make use of geodemographic classifications are having to familiarise themselves with new and updated segmentation systems. The release of these classifications is therefore an appropriate moment to review the current ‘state of the art’ in this particular form of consumer segmentation. This review starts with a brief account of how the marketing applications to which these segmentation tools have been put have evolved since their introduction in 1979, and an assessment of their relevance and scope in relation to other sources of data that marketers can use to segment prospects and customers. The following section of the paper challenges the assumption that the finer the geographic granularity of the units being classified the better the predictive power of the classification, and questions the popular view that geodemographics is useful primarily in contexts where demographic data at the household level are unavailable. This paper presents evidence to suggest that neighbourhood often contributes incremental predictive power in behavioural models over and beyond individual-level characteristics. It is suggested, however, that there is no optimal scale for classifying neighbourhoods. Consumer behaviour within some product categories is better predicted using demographic data for areas more geographically extensive than Census output areas, while for others the appropriate granularity is as low as unit postcodes. The paper then advances a basis for judging which level of granularity, fine or small, is likely to be most valuable for predicting usage levels in different product categories. The concluding sections explain how changes in the marketing applications to which geodemographic classifications are now put are affecting the manner in which these systems are now constructed, resulting for instance in the use of data from sources additional to the Census and in the use of new statistical methods for optimising their predictive performance.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 37 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.