BUS5CA Customer Analytics And Social Media


The project is looking for knowledge and insights related to:

The demographic-based segments and profiles

These are representative profiles of the behavioural characteristics for each segment

The produced segments can be linked to a wider concept of segments in the Australian community.


We have attempted to solve the targeted marketing problem, mostly in the retail sector, with this project.

Bit Bank’s customer list was used for this task.

Clustering has been used to identify different marketing segments and help businesses communicate with their customers effectively (Chen 2014).

Case Study Task 1 – Customer Segmentation Based upon Demographics Data

SAS Enterprise Miner was used for customer clustering.

K-means clustering has been performed with total 5 clusters.

Clusters were made using demographic variables.

This was done in two different ways (Iaci and Singh 2012; Trebuna, et al.

First, the cluster analysis was conducted with demographic variables. Next, it was extended to include the target variable: Subscriber.

Without Outcome Variable:

From the analysis, we have identified 5 clusters.

Take a moment to zoom in if you are unsure of the results.

We are looking at the distribution of each variable among the 5 clusters in the plot.

Cluster 1 is made up of people of various ages.

Cluster 1 is not composed of any significant age groups.

However, it is limited to students.

The cluster does not include those with other career options.

It is dominated mostly by those with an unknown level of education and married people.

It couldn’t be directly mapped with any segments from Roy Morgan.

It encompasses people of various ages as well as a mix of different career options.

This segment does not include people with a high school diploma.

This segment is most popular among married couples.

Could not be mapped directly with any segments from Roy Morgan.

It also features a mixed distribution of age and career.

These variables aren’t significant in this cluster.

The group is comprised of more people who have completed secondary education, and they are all divorced.

This segment is extremely interesting.

These segments can be connected to the “Something Better” segment by Roy Morgan.

They are often divorced and educated.

This age group is dominated by people in their younger years, ranging between 27-37 and having different career options.

For most people, education is second-class. The marital status of the majority of people is half married and half single.

This can be used to map the young optimism segments of Roy Morgan.

This group is associated with optimism and a desire to improve one’s life prospects.

This cluster contains people from different career and age backgrounds.

However, it is home to more people who are married or have never divorced.

Could not be mapped directly with any segments from Roy Morgan.

Note: Not all clusters can be mapped into Roy Morgan Segments.

This is due to the fact that the attributes defined by Roy Morgan Segments are in the current data.

Distribution of each variable within the 5 groups with target variable

The outcome variable has 3 key segments.

Segment 1 and 4, which only include people who haven’t subscribed, are the key segments.

Segment 2 consists of people who have subscribed.

Next, we will analyze whether any demographic variable plays a significant role when defining the cluster.

By examining each cluster using the various demographic variables, it will be possible to identify any variables which play an important role in the cluster’s definition.

We couldn’t find any variables as such in the above graph.

It is clear that there are no distinct segments for customers who have term deposit subscribers and those who don’t.

Case Study Task 2 Customer Segmentation Based upon Behavioural Information

The clusters were only created by using behavioural variables.

It was done in two ways (Seiler, et al.

First, cluster analysis was performed with only behavioral variables. Next, it was run with the target variable Subscription.

Notice: All three behavioural variables are binary

Segments 1 and 2 are the key ones that are created based on default credit & personal loans.

Segment 1 includes people who have not defaulted on a personal loan.

Segment 2 includes people who have personal loan and mortgage loan but have not defaulted.

Segment 5 consists of people who have a personal loan but not a mortgage loan and who have not defaulted.

Segment 1 is the most popular segment, based on the behavioural variable.

It includes people who do not have a personal loan or have not defaulted.

Cross Cluster Analysis – Demographics To Behavioral Segments

Note: This was done using the pivot tables in excel.

The excel workbook is available for further reference.

Based on the clusters created in task 1, we have assigned each person to a cluster.

Excel was used to create a cross tab that included both the clusters.

Visually, the cross tab didn’t show any association between clusters.

To determine the relationship between the outcomes and the combined clusters we would need to examine the number subscribed clusters that fall within each cell.

We would then examine how many subscribers would be in the combined section.

We can see the percentages of subscribers belonging to each cluster in the cross tab.

The cluster 1 of behavioural clusters and cluster 3 of demographic clusters have 113.44% of subscribers.

The lift ratio for each segment is 13%.

Similar to the average of all segments, the cluster consisting of 2 in demography & 5, in behavioural, has 117.86% subscriptions.

All segments marked with green indicate key segments.

Case Study Task 4 Customer Segmentation Based upon Combined Demographic & Behavioural Data

If the cluster algorithm is run using both demographic and behavioral variables.

Maximum clusters have been set at 5.

We have given the maximum clusters as 5.

Segment 5 is the key segment. It includes students with no personal loan.

Segment 3 also is important as it contains all individuals who have a mortgage but have not defaulted.

This is a great target segment for businesses.

Now, run the cluster analysis with the target variable.

Only customers who have subscribed are included in segment 1.

Segment 1 contains people of mixed age and people with mixed careers.

Subscribers don’t have to take out a personal loan.

Personal loan and default credit are the most important variables.



UNLV Gaming Research & Review Journa (16(1)), pp.21-43.

The impact of socio?demographic variables and customer satisfaction on loyalty in the private bank industry.

International Journal of Bank Marketing, 31, pp. 235-258.

The importance and role of standardization and normalization in clustering.

Leave a Reply

Your email address will not be published.