Case Scope: 

•This is a case study from my Analytics Techniques Class

•We were encouraged to explore different models in regression and classification to arrive at best predictors to customer churning behavior

context

•QWE, Inc. is a subscription-based platform for small and medium-sized business to manage their online presence. Using 2-month customer data, the company hopes to develop a proactive approach to better predict customers who are likely to churn in the future.

​

 

Our Approach:

•First, we used the Decision tree and Logistic Regression Model to classify customers into various segments and predict their probability of leaving QWE. These two approaches help us examine factors that contribute most to the likelihood of churn rate. Because the Logistic Regression Model has a higher true positive rate than the Decision Tree, we choose to use the Logistic Regression Model as our primary method to segment our customers.

​

•The top three factors that most influence the customer churning rate are: How long customers have been with QWE, how long it has been since customer’s last login, customer happiness level with our service.

 

•Our analysis leads to the conclusion that if there is limited resources or time, QWE should concentrate on current customer happiness level because QWE is ultimately in the business of providing confidence to the small and medium-sized business of managing their own online presence. As long as the customers are satisfied with the service, they will stay with QWE.

 

Please click here for the full report