BUS5CA Customer Analytics and Social Media

Introduction

Customer segmentation is a critical process that helps businesses tailor their marketing strategies by dividing a broad customer base into smaller, more manageable groups with similar characteristics (Tam et al., 2021). This basically enables an organisation to understand the needs, preferences, and behaviors of its customers, which helps retain customers and ensures better business growth.

Big data at present seems to be an overriding technology; hence, machine learning techniques like clustering are immensely useful to perform these kinds of segmentation tasks (Auriemma Citarella, 2022).  KMeans enables unsupervised learning for similar characteristic classes in a dataset to view better viewpoints for more personal engagement and a product offering (Tabianan, et al., 2022).

The importance of customer segmentation is that it allows uncovering hidden patterns across the customer base, which could be useful in decision-making. Indeed, research shows that focused marketing campaigns toward segmented groups tend to yield better conversion results and higher levels of customer satisfaction (Uddin et al., 2024).

Segmentation will also help in the complete view of customer profiles based on demographic data as well as behavioral data, hence enabling firms to design effective retention strategies. Competitive advantage is driven by the applications of customer analytics, particularly in industries such as financial services and e-commerce, since emerging expectations will be present there Nagarathinam et al., 2024).

This report presents in-depth customer segmentation analysis for a financial institution with demographic, behavioral, and combined clustering analysis, aiming to identify key customer profiles. The nature of the current data set bears relation to three critical questions: customer attrition, retention, and engagement.

Cross-cluster analysis and feature importance emanating from activities undertaken in this study will attempt to bring forth intrinsic patterns that could assist in designing targeted marketing and customer retention strategies.(BUS5CA Customer Analytics and Social Media)

1 Task 1: Customer segmentation based on demographics data

1.1 Demographic segments and profiling

In the process of the demographic analysis, demographic variables were considered including Customer Age, Gender, Marital Status and Education Level. By applying the KMeans clustering algorithm, we segmented the customer base into distinct demographic groups, as presented in Table 1: Key segments of the demographic clusters (CDCs).

The following table represents standardized cluster centers as well as the manner that each demographic feature differs from the other across the five clusters. The values in the table are indicating how each demographic characteristic is distributed across the clusters in relation to global mean, here we see that the values greater than zero mean that this particular feature in this cluster is better or above mean while values below zero means that this feature is below mean in this cluster.

The variation in these characteristics across the clusters can be observed, indicating the heterogeneity of the demographic attributes of the customers and their propensity to purchase goods and services of Swedish industries and firms.

Every cluster in the demographic segmentation is a separate profile, which can give better information about the customers. For example, Cluster 0 consists of the customer segments that are older in age, and the segments include fewer women and those with less education. Whereas, Cluster 1 has relatively more proportion of married women of average age and education level.

The percentage of participants within Cluster 2 with higher education level is higher while single males are more represented in this cluster a bit more than in the other clusters. Cluster 3 is similar in gender distribution to cluster 1 but includes more single individuals. Last, Cluster 4 is significantly younger and has a slightly higher percentage of marriage.

(BUS5CA Customer Analytics and Social Media)

This demographic segmentation gives a clear vision of the major customer segment and the best approach that can be used to address their needs. Each of the clusters discussed features specific properties that can be utilized to improve the level of user interaction, loyalty, and satisfaction.

For instance, the correspondence of findings from Cluster 1, which consists of married females, could be used as the basis for targeted promotions, or the results obtained in Cluster 2, which includes single, educated males, could be used to create specific offers.

1.2 Variable importance

This study analyzed feature importance with an aim of identifying which of the demographic variables have to do most with the differentiation between existing customers and the churned ones. Using a decision tree classifier, the relative importance of each variable was calculated.

Looking at the table below, one can notice that the greatest contribution has been obtained for the variable: Customer Age equal to 42.23 percent on the total of the prediction model. This is trailed by Education Level, which accounts for, 31.67%, it would therefore betray personality type and these two as the most influential places for predicting customer behavior.

Gender and Marital Status have lower importance values of 13.59 % and 12.51 % respectively but still influences the decision making on health care services significantly

Feature Importance for Existing Customers can be a useful tool in gaining an understanding of the importance of each of the major demographic features. By looking at the bar chart in figure 2 above, it is so evident that Customer Age has the highest mean implication to customers’ commitment to the company with Education Level being the second most influential factor.

What this circumstance means is that older and more educated customers could prove more faithful to the company. Analysing the results found that AOV, Education, Gender, and Marital Status are significant factors and when looking at Customer Attrition these four scale are also important but the significance of these four scale shows that Gender and Marital Status are not as significantly important for determining customer attrition as much as Education or Age.

The outcomes of the feature importance analysis are beneficial when it comes to the construction of customer retention plans. Looking at the current demographic variables it is possible to determine age and education to engage and retain the most influential customers at the company.

1.3 Differences in segments between subscribed  and non-subscribed customers

The demographic segmentation analysis revealed notable differences between existing and attrited customers, as shown in Table 4: Table 4: Cluster Centers for Existing Customers (Demographic Segments) and Table 5: Cluster Centers for Attrited Customers (Demographic Segments). The above tables display the coordinates for both clusters, as well as all customers by variables like the Customer Age, Gender, Marital Status, and Education Level.

Disparities between Existing and Attrited Demographic Cluster Centers. Most importantly, Customer Age revealed the greatest difference between Clusters 0 and 2, an observation suggesting that only those who are old subscribe to the service. Moreover, there is significant variability in the distribution of Education Level in Clusters 0 and 4, which indicates that higher education levels might be associated with retention, especially in Cluster 0. These insights enable the understanding of how demographic aspects affect customer behavior and their subsequent loyalty.(BUS5CA Customer Analytics and Social Media)

2 Task 2: Customer segmentation based on behavioural data

2.1 Behavioural segments and profiling

The result of the behavioural segment is shown in the form of 5 clusters which have some related variables like credit limit, revolving balance, average open to buy amount, transaction amount and the amount of utilization ratio. Revolving credit customers in cluster 0 are those with low average revolving balance and revolving utilization ratio what depicts prudent spending. Cluster 1 has high transactions and credit lines but average credit usage; therefore, vendors active but not frequently charging their credit limits.

Cluster 2 has the lowest credit limit and most similar open to buy but the highest of each utilization ratio which shows a riskier behavior. Cluster 3 is made up of customers who transacted frequently but have low credit transaction levels and high utilization ratios indicating that they rely on credit heavily.

Finally, the Cluster 4 reflecting high credit limits and low pattern of usage, indicate the presence of affluent customers who possess large credit line, but they seldom utilize it. These clusters help understand a range of behaviours displayed by customers.

The behavioral segmentation reveals five distinct clusters with unique profiles based on various financial and transaction features as shown in Table 8: Behavioral cluster Profile. Based on the results of the analysis, customers from Cluster 0 are characterized by lower credit limit and revolving balances with below average ratio of credit utilization.

The elements that define Cluster 1 include greater transaction value and volume, and a relatively high credit limit and a somewhat lower use rate. This cluster has lower limit for credit and revolving balance but has positive utilization ratio which mean credit is frequently used.

In this case Cluster 3 shows lower transaction tendency as reflected by the transaction amounts and count and low credit utilization suggesting lesser financial activity. The last, Cluster 4, has high credit limits, open-to-buy balances, and the percent of new volume with little dependency on revolve balance, proving that the customers are more financially sound. Each cluster is useful in understanding customers’ financial behaviour in order to allow for a targeted approach.

2.2 Variable importance

The important variables based on each behavioral segment targeting existing customers, as seen in Table 9: Given in the Features Importance (behavioral Variables) and Figure 5, the Feature Importance for Predicting Existing Customers show that ‘Total_Trans_Amt’ total transaction amount has the highest feature importance percentage of 23.37%, while the ‘Total_Trans_Ct’, total transaction count has a percentage of 20.29%.

Table 9: Feature Importance (Behavioral Variables)

These two variables are extremely important for customer retention, meaning that the customers who make more transactions of higher amounts and of greater frequency will likely to remain loyal. Other important variables are Total_Revolving_Bal which represents credit balances (13.95%) as well as Total_Ct_Chng_Q4_Q1 and Total_Amt_Chng_Q4_Q1 which highlight the changes in transaction behaviour in Q1- Q4 (12.75% and 9.89% respectively). The comparative importance of Avg_Utilization_Ratio with the Credit_Limit and Avg_Open_To_Buy it is considerably lower but still sharply revealed.

(BUS5CA Customer Analytics and Social Media)

2.3 Differences in segments between subscribed  and non-subscribed customers

The comparison of the clusters based on the behavior of existing and attrited customers is presented below in Table 10 (Cluster Centers for Existing Customers), and Table 11 (Cluster Centers for Attrited Customers).

Table 10: Cluster Centers for Existing Customers (Behavioral Segments)

Similar to the findings described in Cluster 1, for existing customers, Cluster 2 has higher Total_Trans_Amt ( 2.64) and Total_Trans_Ct (1.75) implying that these customers are active participants in the transactions and engage in spend heavily. Cluster 1 remains moderately active with slightly lower average of Total_Amt_Chng_Q4_Q1 (-0.17) which suggest that activity here is less erratic than in Cluster 2. Analyzing the characteristics of Cluster 3 we see that the values decrease in the Total_Trans_Ct and Avg_Utilization_Ratio variables, which stresses the group is less active and potentially closer to dormancy.

Table 11: Cluster Centers for Attrited Customers (Behavioral Segments)

For attrited customers, the trends are different. Looking at the preventive control variables, cluster 3 was characterised with a Credit_Limit of 2.26 and Avg_Open_To_Buy of 2.27 meaning that the customers had a lot of credit available but little activity on them thus the tendency to attrite. Hmmm… In fact, Cluster 0 appears to consist of a lesser active customer base, and as I have noticed a low Credit_Limit (-0.26) and overall negative transaction-related values which suggest high potential of churn out of the customers.

Figure 6: Cluster Centers for Existing Customers and Attrited Customers (Behavioral Segments)

These disparities are well illustrated in figure 6 below. Comparing with attrited customers, the behavioral features specific to Avg_Utilization_Ratio and Total_Trans_Ct are lower for all the variables than with existing customers, which underlines a significant difference in activity in customer attraction or churn.(BUS5CA Customer Analytics and Social Media)

3 Task 3: Cross cluster analysis – demographics to behavioural segments

In this task, cross-cluster-analysis is conducted on the demographic and behavioral segmentation variables with a view of establishing how they relate with customer attrition. As executing the use of pivot table, putting demographic cluster into rows and the behavioral clusters into columns was made. This made it possible for the researchers to adequately examine customer distributions along the two dimensions.

3.1 Associations between the two types of segments

To test for relationship between demographic and behavioral clusters, a chi-square test of independence was used. Therefore, it is evident that demographic clusters are statistically significant with behavioral segments, establishing the chi-square statistic of 833.3 for the 0 p-value. This has emphasised the fact that in customer segmentation, demographic features as well as behaviours should be considered.

3.2 Relationship between the outcome (Attrition flag) and the combined demographics and behavioural segments

Some details can be examined in the figure 7 below, where additional statistical calculations were made to determine the correlations between the segments combined with the outcome variable, Attrition Flag. Subscription percentage was then computed for each possible demographic and behavioral clusters pair at the application level.

The results thus suggest that, for example, demographic segment 3 paired with behavioral segment 0 yields the highest subscription percentage of 93.28%, as well as demographic segment 4 with behavioral segment 3 with a subscription percentage of 97.67%. These segments are the most valuable as they are likely the most loyal, whereas demographics 0, behavioral segment 2 (49.18%) indicates that it is most vulnerable to attrition based on subscription rate.

The results presented here also show the relationship between the customer behavior, customer demographics, and their customer loyalty. Knowledge of such relationships helps the companies to do a better job in identifying the high risk segmentation and to enhance the retention functions.

4. Task 4: Customer segmentation based on combined demographic and behavioural data

4.1 Key segments for the whole dataset

In Task 4, the customer segmentation was done with both the demographic as well as some behavioral data. The idea was to provide a comprehensive view of the consumer groups as a result, rather than approaching the problem in two different steps, namely demographic and behavioural analysis. In total, five significant customer segments were determined by applying the KMeans clustering methodology to the specification of demographic and behavioral attributes.

The clustering results are shown in Table 14: Cluster Centers for Combined Demographic and Behavioral Data, giving the standardised values of each feature in the clusters. Because of this arrangement, customer segment is captured more comprehensively in terms of strategizing for the business. Combined Demographic and Behavioral Segments – Cluster Centers enables one to understand the patterns within the demographic and behavioral characteristics within the clusters and how they inform different customer groups.

Cluster 0 shows the customers with credit limit below the overall average by 6% with slightly lower utilization ratios which indicates lesser credit spending. Fewer numbers of transactions and a low level of revolving balance can also be seen for these customers, which shows that they are not very active financially.

Customers in Cluster 1 have the least gender balance and high credit use, clustered together. These customers display moderate amounts of revolving balances and transactional volumes on their credit cards while they are comparatively higher in terms of risk than the former clients. Customers in cluster 2 are those who can easily afford what they want because they have higher credit limits and have relatively high spending power.

These people show a greater availability to open to buy credit yet less expenditure in terms of actual transactions as compared to other segments.

Cluster 3 represented the highest transactional level. This customers’ segment is characterized by high frequency of transactions and the largest total revolving balances. Credit limits of their cards are reasonable, however they are equally wise in their choices within this aspect of utilization.

Finally, the last cluster is characterized by younger customers who demonstrate a high use of credit and moderate transaction volume. They average credit limits while their revolving balances are somewhat lower and they signify moderate utilization rate.

4.2 Important Variables Considering the Outcome (BUS5CA Customer Analytics and Social Media)

When considering the outcome (Target: In the case of the (Existing Customer), the content of the variable of priorities is primarily behavioral. Out of all the features Total Transaction Amount and Total Transaction Count have the highest importance value of 0.22 and 0.19 respectively.

They seriously affect the orientation of existing customer accounts and imply that transaction activity is at the heart of customer loyalty. The Total Revolving Balance (H2) and the Total change in Transaction count (Q4-Q1) are quite noticeable with the importance values of 0.12 and 0.12 respectively; this means customers, with high revolving balances and variation counts in transaction, are less likely to churn firms.

On the other hand, the values of Demographic variables such as Customer Age and Gender are assigned lower importance, which implies that customer retention is not much affected by the demographics of customers rather, it is more of a behavioral aspect. Credit Limit, Education Level, and Marital Status are small factors but are also important in a way to help explain the current behavior of the customers.(BUS5CA Customer Analytics and Social Media)

4.3 Different segments and profiles identified (BUS5CA Customer Analytics and Social Media)

Yes, there are significant differences between the segments and profiles found in Task 3 and Task 4. In Task 3, a cross-cluster analysis was conducted between the demographic and behavioral segments, whereas in Task 4, the entire data set was clustered based on the demographic and behavioural data in a direct manner.

In Cluster Distribution – Task 3 (Demographic & Behavioral Combined), the result of the segmentation was more specific and spread out in different categories of demographic and behavioral elements (such as 4-0, 3-0, 2-1). Some of the dispersal clusters like 4-0 had relatively higher counts of 700, while some other dispersal clusters like 4-1 had smaller count of 60. This indicates that Task 3 led to discovering more nuanced clusters for the possible combination of demographic and behavioral attributes.

Cluster Distribution – Task 4 (Combined Demographic & Behavioral): This time the number of large clusters is quite limited at five. Cluster 1 has the highest number of customers (2508) followed by Cluster 0 (2024), whereas Cluster 3 has the least count (680). Comparing this result to Task 4, it can be concluded that combining demographic and behavioral data provided for segmentation into fewer but more general clusters.

Looking at Table 19: In Cluster Centers for Combined Demographic and Behavioral Data (Task 4) the profiles are more general and depict major tendencies in the customer base. For example, while the Task 3 carries out the inspection on Cluster 3, where the revolving balances, transaction amounts, and the sizes of clusters are different, which means that they are somehow different in terms of engagement level.

Conclusion

This report has gone ahead to conduct customer segmentation based on demographic, behavioral, and combined data in identifying key customer profiles and their trend of behavior. KMeans clustering brought up some distinct patterns in the segmentation analysis, which may be helpful in devising more specific marketing strategies in terms of better customer retention.

From Task 1 targeting demographic segmentation emerged two most decisive variables of age and education for determining customer loyalty. The second task was related to behavioral segmentation, in which transaction-related features are proved to be most important about predicting customer retention regarding the total amount and count of transactions.

The cross-cluster analysis that was performed in Task 3 revealed some associations of demographic and behavioral clusters, which means that both of these types of data bear great importance for any explanation of customer attrition. Task 4 applied combined demographic and behavioral clustering to develop more generalized customer profiles, showing clearer patterns of behavior and scopes for more strategic planning in customer engagement.

The in-depth analysis justifies the fact that integration of both demographic and behavioral insights leads to better customer segmentation and thus more effective retention strategies across segments. Companies will be able to leverage their resources to undertake proper initiatives that provide them with means to maintain a relationship with their customers in the long term, once they gain insight into the most critical drivers of customer loyalty.(BUS5CA Customer Analytics and Social Media)

If you’re looking to deepen your understanding of BUS5CA Customer Analytics and Social Media, assignment writing services can help you craft well-researched assignments that cover key topics like customer behavior and social media trends. For expert assistance, visit “Assignment Writer”. Also, for those seeking tools to improve their academic performance, check out “Check My Assignment” to easily analyze your assignments.

Reference

Auriemma Citarella, A. (2022). Application of machine learning techniques to biological big data.

Nagarathinam, A., Chellasamy, A., & Rangasamy, S. (2024). Strategic Data Analytics for Sustainable Competitive Advantage. In Data-Driven Decision Making (pp. 77-106). Singapore: Springer Nature Singapore.

Tabianan, K., Velu, S., & Ravi, V. (2022). K-means clustering approach for intelligent customer segmentation using customer purchase behavior data. Sustainability, 14(12), 7243.(BUS5CA Customer Analytics and Social Media)

Tam, P. T., Son, D. M., Le Tan, T., & Ha, H. (2021). Data Driven Customer Segmentation for Vietnamese SMEs in Big Data Era. Macro Management & Public Policies, 3(2), 33-43.

Uddin, M. A., Talukder, M. A., Ahmed, M. R., Khraisat, A., Alazab, A., Islam, M. M., … & Jibon, F. A. (2024). Data-driven strategies for digital native market segmentation using clustering. International Journal of Cognitive Computing in Engineering, 5, 178-191.(BUS5CA Customer Analytics and Social Media)