Segmentation/Cluster Analysis

Main aspects

  • Groups together respondents with similar characteristics
  • Each respondent belongs to only one group
  • Via hierarchical, non-hierarchical or latent class methods
  • Again commonly used on attitudinal data e.g. two dozen statements concerning prescribing may be used as input to a cluster analysis which might identify three or four different GP 'types'
  • Number of clusters is determined by how much variation they explain and their interpretability
  • Each respondent can be coded with the relevant cluster number and these can be used as breaks in crosstabs
  • Resultant clusters often targeted in ad campaigns