Home / Summaries / Marketing Models / clusters-customer-cluster

Market segmentation and cluster analysis

Q: What is k-means clustering?

It is not hierarchical but it is a partitioning technique . Partition is the description of which customer belongs in which cluster , for as many clusters as deemed appropriate . Tell the model you want e.g. 3 clusters . For each amount of clusters you have to run separate models.

Q: What are the steps in interpreting the clusters?

Examining the means of each variable for each segment. Provides a profile as to how the customer segments vary . Creating a table or plot of 95% CI intervals for all k segments by all p variables tells which differences are significant. Or ANOVA . Compare segments on customer data that were not used to derive clusters .

3 important questions on Market segmentation and cluster analysis

What is k-means clustering?

It is not hierarchical but it is a partitioning technique. Partition is the description of which customer belongs in which cluster, for as many clusters as deemed appropriate.
Tell the model you want e.g. 3 clusters. For each amount of clusters you have to run separate models.

How does k-means clustering work?

Tell the computer how many clusters you want.
The computer randomly assigns each customer to a cluster.
Then centroids are calculated for each of the variables and for each cluster.
Distance of customers to centroids is calculated and if the distance of the customer to another centroid is smaller, he/she is added to that cluster.

What are the steps in interpreting the clusters?

Examining the means of each variable for each segment. Provides a profile as to how the customer segments vary.
Creating a table or plot of 95% CI intervals for all k segments by all p variables tells which differences are significant. Or ANOVA.
Compare segments on customer data that were not used to derive clusters.