Segmentation is a structured and iterative process to group objects – customers, accounts or transactions – into similar segment.

The objects –customers, accounts or transactions – within a segment will be similar to each other and will be different to that of other segments

The similarity of the objectives can be defined based on some attributes.  For example, a simple segmentation can be of segmenting customers into male and female segments based on gender.  Most of the time, multiple variables /attributes will be used to create relevant segments.

Segments can be used to understand customer needs, transactional behavior, response to an offer, or risk behavior. The relevance of the segments will depend heavily on the context and objective of the segmentation.

Segmentation can be build using Objective Segmentation Methods such as Decision Tree and Subjective Segmentation methods such as K-Means Clustering

Transactional Segmentation

Based on customer transactions- credit and debit transactions on bank saving accounts or spend transactions at a retailer – segments can be created to understand customer transactional behavior.

Transaction segmentation can be built using transaction frequency, value and category.  Since there is no defined rule around the segments, subjective segmentation method K-means clustering will be used to group customers based on their transactional behavior.

In K Means clustering, similarity measure can be defined using Euclidian Distance Measure. There are other similarity measures as well. K Means clustering technique requires numeric variables as input variables. Also, K Means technique requires numbers of clusters as an input parameter.  Iteratively the objects – customers or transactions – will be grouped together using K-Means clustering algorithm.

Transactioan Segments-an example

Transactioan Segments-an example

Attached graph illustrates 3 segments- Internet Only, Brick-Mortar Only, and Highly Engaged – created using customer spend across merchant categories.

Similarly based on customer demographic information Customer Life Stage Segmentation can be built. Also, based on customer life style information Customer Life Style Segmentation can be built and used for various objectives.

Subjective Segmentation methods such as K-Means Clustering

Building Subjective Segmentation using K Means Clustering on Transcript Data