Customer Segmentation

  • Identify and classify the total target population into ‘mutually exclusive and collectively exhaustive’ segments or groups.
  • Enables Targeted Marketing Activities:
    • Maximize M-ROI: by classifying segments based on potential (both size and relevance), demographic profile and media habits.
  • Enables Tactical Resource Allocation:
    • Product Planning: by determining what price different groups of consumers are willing to pay or can afford. Segmentation allows you to serve each segment at a price level its members can afford. For e.g. different vehicle trims for different segments.

A Note on Factor Analysis

  • What does Factor Analysis do?
    • It reduces a larger set of variables into a smaller set of ‘artificial’ variables (also called principal components) that account for most of the variance in the original variables. Factor analyses is related to PCA, though the two are not exactly identical.
    • In statistical terms, Factor analysis is a method used to describe variability among observed, correlated variables in terms of a potentially lower number of unobserved variables called factors.
    • Since it condenses information in to manageable themes and classes, it is also a potent “data reduction” technique.
  • How does ‘Factor Analysis’ helps to make our analysis procedure more robust and statistically valid?
    • Minimizes multi-collinearity between variables
    • Summarizes sentiment in to themes that are similar
    • Aides in statistical validity of regression or for that matter any advanced modeling

A Note on Latent Class Analysis

  • Clustering Algorithm
    • Latent Class Analysis (LCA) is a statistical method for finding subtypes of related cases (latent classes) from multivariate categorical data.
    • It is an “improved” cluster analysis, which uses statistical methodology rather than mathematical to construct the results.
    • It is based on statistical concept of maximum likelihood (ML).
    • The main difference is that cases are not absolutely assigned to classes, but have a probability of membership for each class.
    • Data type is not a limitation: it can deal with all types of data – binary, continuous and categorical unlike most traditional methods.
  • How does ‘LCA’ helps to make our analysis procedure more robust and statistically valid?
    • LCA decides the best number of classes based on two statistics:
      • AIC – Akaike Information Criteria
      • BIC – Bayesian Information Criteria
    • Chooses the number of classes which minimizes AIC, BIC

Email to someoneShare on LinkedInGoogle+Tweet about this on TwitterShare on FacebookPrint this page