Implementing a Customer Lifetime Value Predictive Model: Use Case

Since predictive algorithms are really just mathematical formulas that can be applied to many different problems, many organizations have a difficult time understanding how they can be applied and implemented within their existing Business Intelligence environment. Fortunately, SAP provides several tools that offer an end-to-end solution for generating and visualizing analysis sets, fitting predictive models, and implementing those models into the BI platform. This example shows a case study of an online retailer to demonstrate fitting and implementing a predictive model for customer lifetime value.

Background

AdventureWorks Cycle Company is an online retailer of bicycle supplies and parts. Through their e-commerce channel, they sell to individual customers around the world. AWCC sells a wide range of products, including both inexpensive parts and complete bicycles worth thousands of dollars. They have already acknowledged a large variance in the value of an individual customer, from low value one-time parts customers, to very high value repeat purchase customers. Therefore, AdventureWorks would like to devise a lifetime value model that takes into account a few simple variables to return an estimate of the customer’s value potential.

Building a Customer Lifetime Value Model

To start this project, we’ve estimated and imported into SAP Predictive Analysis the lifetime value of a set  of customers with extensive purchase history and a few factors we have available on newer customers: age, income, and an indicator of whether they purchased a specific product (in this case, we’ll just name it Product X). This information could be from a spreadsheet or a Web Intelligence report in of the existing BusinessObjects reporting environment.

____________________________________________________________________________________________________________________________________

Figure 1: CLV Input Data

We then bring the R-Multiple Linear Regression algorithm into the predictive workflow on the Predictive pane, since the value we want to predict (Value) is continuous.

Figure 2: Predictive Workflow

Then, configure the R-Multiple Linear Regression component with independent columns Age, Income, and ProductX to predict the dependent column Value. We’ll also save our model as CLVLRModel.

To download full PDF and Continue Reading…

About Hillary Bliss
Hillary Bliss is the Analytics Practice Lead at Decision First Technologies, and specializes in data warehouse design, ETL development, statistical analysis, and predictive modeling. She works with clients and vendors to integrate business analysis and predictive modeling solutions into the organizational data warehouse and business intelligence environments based on their specific operational and strategic business needs. She has a master’s degree in statistics and an MBA from Georgia Tech.