Imagine a company with multiple lines of revenues coming from diferent products, but all customer can access these different products through the same account and the same online platform. My goal is to predict the churn for each customer.
Should I perform customer segmentation into clusters and build a churn prediction model for each segment?
The straight foward path would be to get all relevant features for all customers and try to predict the churn for all of them. The different bahaviors and personae will theoretically reflect on the features.
What I do expect to get from the clustering are the personae this company intends to serve.
In this case, a churn will be defined as a custumer who stops consuming any product or close its account.
My objective here is to build a churn prediction model according to each pearsonae behavior pattern and I expect from this approach to have better features for each because now we understand them more and then more acurate models.
This would also lead to more personalized business strategies to avoind the churn.
But I don’t know if this is a good practice and if I should try the straight foward approach or a different one.