Ramgopal Prajapat:

Learnings and Views

K Means Clustering Examples and Practical Applications

By: Ram on Jan 07, 2021

Price -Segmentation v1Pricing Segmentation: E-retailers or e-commerce companies have taken the retail industry by storm. They are offering luring offers and discounts. They aim to move from discount led to convenience or differentiation led offering over a period in time. Some of them have been forced to start the journey.

One of the large retailer wanted to segment the customers based on customer spend patterns and understand price sensitivity of the customers. Some of the segmentation variables considered are – total spend, value of discounts, % discounts across transactions, number of items bought on discounts etc and used k means clustering to find the discount orientation of the customers.

Lotalty SegmentationLoyalty Segmentation: Every organization wants their customers to be loyal but some time defining loyalty could be a tricky. Some of the common expectations are frequent spend, higher spend, higher ticket size and diversified spend.  A large Indian retailer was looking to understand customers from these 4 dimensions and wanted to develop segments for building a focus engagement strategy for each of the target segment.

Spend Behaviour Segmentation: Developed Customer segmentation for a grocery chain retailer using two year spend pattern on their loyalty card. Some of the key dimension considered for K Means clustering segmentation are Count of Transactions, Value of Transaction, Average Transaction Value and various transformation of these spend engagement indicators.

Branch Geo Segmentation: A large global ATM service provider wanted to define segmentation based on geo-location of ATMs.  ATMs’ Latitude and longitude information were used for creating the clusters. These clusters were aligned to a particular ATM management unit.

Customer Need Segmentation: Based on various customer interest, life stage and product holding information have been used for creating customer need based segmentation. These segments were aligned to different groups of relationship managers to manage these clusters effectively. These customer clusters were different based on their needs, channel of preferences, service expectations.

Category Segmentation: Customer spend patterns vary by merchant category (e.g.  Restaurant Spend, Grocery Spend, Entertainment & travel spend etc). Aim of the segmentation was to create different clusters which are significantly different based on spend across categories.  Then, answer some of these questions.

  • Who are these customers?
  • Why are they behaving the way to?
  • What action can be taken or what treatment strategy should we develop?

Donor Segmentation: For a Non-Profit Organization, a segmentation was to understand level and response of people who will give donation and continue to contribute for a cause.

Some of the dimensions used for K Means clustering were

  • Salary
  • Gender
  • Number of times and Value of contributed in the last 6/12 months
  • Age
  • Participation in Events and Type of Events (e.g. conferences, sports, environment, animal rights, women empowerment etc)

Server Clustering: For a financial services client, aim was to find clusters which should be given for annual contract together.

The details used for the clustering included usage and characteristics of servers. Some of the example of the variables are

  • Server Type
  • Number of requests coming to server in a week/month etc
  • Sever Criticality factors
  • Age of the server
  • Since last maintenance
  • Type of maintenance

Healthcare Fraud Detection Segmentation: For an Indian healthcare provider, aim was to find claims which could be fraudulent. K Means clustering method was used for anomaly detection and claim routing to right claim adjudicator.  Some of the K Means clustering dimensions or variables used were

  • Dieses category
  • Claim Type
  • Age Group
  • Diagnosis Code
  • Hospital Type Or Healthcare Setting

Customer Service Segmentation: For a large Australian Energy provider, the objective was to provide differentiated services to its customer base.  Same segmentation would be used for targeted and differentiated offer and service solicitations. Some of the relevant variables used were

  • Customer Value in last 6/12/18/24 months
  • Customer Type – Individuals and Small Businesses
  • Product type (e.g. Gas, Electricity etc)
  • Length of Relationship
  • Overall consumption
  • Number of complains
  • Number of fines


Leave a comment