[ad_1]
CRM data mining analyzes customer behavior to improve marketing campaigns and increase sales. Descriptive analytics uses segmentation and clustering to group customers by characteristics, while predictive modeling measures correlation between factors to predict future behavior. Specificity is important, and different methods are used, including univariate, CHAID, CART, and multivariate regression models.
CRM (Customer Relationship Management) data mining refers to the process of researching customer relationship databases and analyzing the data collected on customer behavior. This data helps marketers better focus their campaigns, which leads to increased customer retention and sales. CRM data mining is also known as data exploration and knowledge discovery. There are two main categories associated with data mining: descriptive analytics and predictive modeling.
Descriptive analytics uses segmentation and clustering to better analyze a particular pattern of behavior among a particular group of customers. Customers can be grouped by gender, age, race and other categories. The main goal of a segment is to provide the marketer with a group of similar customers in order to more effectively mine data for useful insights.
Clustering aggregates groups of segments. Each cluster is mutually exclusive and is characterized by a set of predetermined characteristics. For example, a cluster might include women aged 18 to 25 who purchased a particular nail polish during the last two weeks of December 2010. This is an example of qualitative CRM data mining.
In nonexclusive segments, another form of descriptive analytics, one particular set of customer behaviors leads to an entirely new set of behaviors. For example, a group of clients might spend a significant amount of money on spa services, but not spend a lot of money on related services such as salon and hair care. This type of CRM data mining requires more advanced statistical analysis than basic segmentation.
Predictive modeling is the more popular of the two CRM data mining categories. Measures the degree of correlation between two factors of customer behavior and the statistical reliability of that correlation. The predictive model is built using a data mining application that assigns scores to each customer, indicating the likelihood that the customer will behave the same way in the future. For example, the model can help a marketer determine the likelihood that a married male customer between the ages of 31 and 42 with children will purchase a particular brand of lawn mower within the next six months.
Specificity is very important in CRM data mining using predictive models. There are different types of methods used for this purpose. A univariate model compares a single variable to several other variables to determine the relationship with the highest correlation. Chi-square automatic interaction detection (CHAID) and classification and regression tree (CART) analysis models show decision trees, where one variable causes one or more variables to instantiate. A multivariate regression model tests several variables against each other to evaluate possible correlations.
[ad_2]