Understanding Customer preferences requires analyzing and processing huge volumes of data. CRM analytics uses data mining to a great extent. CRM analytics can provide customer segmentation groupings (dividing customers into those most and least likely to repurchase a product); profitability analysis (identifying high value customers); personalization (the ability to market to individual customers based on the data collected about them); and so on.

Understanding Customer preferences leads to increased sales and service as the strategies are more targeted. It also contributes to lower costs and more competitive pricing.

While techniques and approaches in CRM Analytics may depend on the actual business problem, let us look at Segmentation and Response Modeling as examples.

Segmentation

It is well known to marketers that not all customers are alike. Retailers are keen to understand the psychological and behavioral differences between potential buyers. They wish to identify geographical segments based on customer satisfaction.

Cluster analysis can be used to identify homogeneous subgroups of respondents. Hierarchical cluster analysis can be used for segmentation. Wald’s method can be used to find the distance between two clusters at each step. Hierarchical tree, called dendrogram can be cut at the stage, where distances between clusters start rapidly increasing. This yields the required segments. Further, profiling of clusters can be done to understand the patterns in each cluster.

Hierarchical tree can be provided indicating the various segments. Profiling output can be provided indicating the behavior within each segment.

Insights can be provided for each cluster, indicating how satisfied the people are, across industries. This would enable more targeted and successful marketing.

Response Modeling

Response Models can help us unlock some of the typical problems associated with Churn in Telecom companies.

Service providing companies are concerned over the large scale churn affecting their growth. It is of prime interest to companies to increase their subscriber base and maintain their existing customers.

Exploratory data analysis can be done to identify the patterns in data and suitable transformed variables can be created. Logistic regression can be used to estimate the probability of churn. Split-sample validation can be used to ensure good performance of the model. A prioritized list of customers can be derived based on the propensity to churn. These customers can be targeted for retention campaigns.

Key drivers of churn behavior can be identified. Lift charts can be provided to measure model performance. A score can be provided for each new customer based on his probability to churn.