Bus5Ca Customer Analytics And Social Assessment Answers

Answer:

Introduction:

In this project, we have tried to solve the targeted marketing task, mainly seen in retail sector. For this task we have used the customer database of Bit Bank. Using the clustering techniques, we have tried to identify different marketing segments & help business to have effective communication with their customers(Chen 2014).

Case Study task 1: Customer Segmentation based on demographics data

The customer clustering has been performed using SAS Enterprise Miner. We have run the K-means clustering with total 5 clusters. Only demographic variables has been used for creating the clusters. This has been done in two ways(Iaci & Singh 2012; Trebuna et al. 2014). Firstly, the cluster analysis has been done with only demographic variables & then it was run including the target variable: Subscription.

Without Outcome Variable included:

We have identified the total 5 clusters from the analysis.

Note: Please zoom in if the results are not very clear.

In the above plot we are exploring the distribution of each variables among the 5 clusters.

Cluster 1 consists of fairly mix age group people. There is no significant age group present in cluster 1. But it consists of only people who are students. People with other career options are not included in the cluster. It is dominated by the people with unknown education level & people who are mostly married. Couldn’t be mapped directly to any segments from Roy Morgan.

It includes people of different age as well as mix of different career options. Education level is coming out to be significant for the segments as only people with unknown education level are included in this segment. Most number of people in the segment are married. Couldn’t be mapped directly to any segments from Roy Morgan.

It also has mix distribution of age & career. Both the variables are not significant in this cluster. But higher number of people in the group has secondary level of education & all of them are divorced. This is very interesting segments. People in this segments can be mapped to “Something Better” segment from Roy Morgan. People are educated but divorced which implies they understand their current situation is not working out & wants something better than their current life.

This segment is dominated by younger generation people age ranging from 27-37 but having different career option. The education level for most of the people is secondary & the marital status consist of half married & half single people. This clusters can be mapped to young optimism segments from Roy Morgan. People in this groups are most often associated with high optimism & who are always looking to improve their prospects of life.

This cluster includes mix of people of different age, career & education level. But it has higher people who are married & no divorcee. Couldn’t be mapped directly to any segments from Roy Morgan.

Note: All the clusters couldn’t be mapped to Roy Morgan Segments. This is because the attributes defined in Roy Morgan Segments are present in the current data set.

Distribution of each variables among the 5 clusters with target variable

We have identified 3 key segments from the outcome variable. Segment 1 & 4 consist of only people who have not subscribed. The segment 2 consist of people who have subscribed. Now, we will examine whether any demographic variable have significant roles in defining the cluster.

If we explore each cluster based on the different demographic variables, we will be able to find any variable which have significant role in defining the cluster. But from the above plot, we couldn’t find any variable as such. So, there is no clear differences in segments for customers subscribed to term deposit and those who did not.

Case Study task 2: Customer segmentation based on behavioural data 

Only behavioural variables has been used for creating the clusters. This has been done in two ways(Seiler et al. 2013; Davis et al. 2008). Firstly, the cluster analysis has been done with only behavioural variables & then it was run including the target variable: Subscription.

Note: all 3 behavioural variables are binary

The key segments that are created on basis of default credit & personal loan. Segment 1 consist of people who don’t have personal loan & not defaulted. Segment 2 consist of people with personal loan & mortgage loan but has not defaulted yet. Segment 5 consist of people who has personal loan but don’t have mortgage loan & has not defaulted yet.

Based on the behavioural variable, we have identified segment 1 as key segment as this consist most number of people who have taken the subscription. It consist of people who doesn’t have personal loan & has not defaulted yet.   

Case Study task 3: Cross cluster analysis – demographics to behavioural segments

Note: This has been done in excel using the pivot table. For further reference kindly refer to the excel workbook provided.

Based on the clusters formed in task 1 for behavioural & demographic segments, we have assigned each individual to clusters. Using excel, we have created a cross tab of both the clusters. Visually examining the cross tab, we couldn’t see any association between the clusters formed.

To understand the relationship between the outcome & the combined clusters, we would examine the number of subscribed clusters falling into each cells. We would examine number of subscriber who would fall in the combined section.

In the above cross tab, we are observing the percentage of subscribers falling into each clusters. For example, in cluster 1 of behavioural clusters & cluster 3 of demographic clusters there are 113.44% subscriber. The lift ratio of 13% in the segment. Similarly, the cluster formed from 2 in demography & 5 in behavioural consists 117.86% subscriber as compared to average of each selected segments. All the segments marked with green are key segments for the business. 

Case Study task 4: Customer segmentation based on combined demographic and behavioural data

When the cluster algorithm is run with both demographic & behavioural variables. We have given the maximum clusters as 5. Most significant variables is default credit as each 5 segments have people who have either default history or has not defaulted yet. The key segment is segment 5 which consist of only students who doesn’t have personal loan. Similarly segment 3 is the key segment as it consist of all people who have mortgage loan but has not defaulted yet. This can be major target segment for business.

Now running the cluster analysis with the target variables. In segment 1, only customer with subscribed people are included. In segment 1 there are mix age of people, also consist of people with mix career & who has not defaulted yet. People who have subscribed have no personal loan. So, the most important variable is personal loan & default credit.

References

Chen, J., 2014. Retail Customer Segmentation using SAS,

Davis, A., Gunderson, M. & Brown, M., 2008. THE EFFECT DEMOGRAPHICS HAVE ON THE DEMAND, Florida.

Iaci, R. & Singh, A.K., 2012. Clustering high dimensional sparse casino player tracking datasets. UNLV Gaming Research & Review Journa, 16(1), pp.21–43.

Seiler, V., Rudolf, M. & Krume, T., 2013. The influence of socio?demographic variables on customer satisfaction and loyalty in the private banking industry. International Journal of Bank Marketing, 31(4), pp.235–258.

Trebuna, P., Halcinova, J. & Fil’o, M., 2014. The importance of normalization and standardization in the process of clustering. IEEE, 12, p.381.