I'm trying to build a simple model to compute Customer Lifetime Value (CLV) based on Recency, Frequency and Monetary (R,F,M). Browsing the web, I've found this question RFM & customer lifetime value modeling in R and after having looked at the answer given, some doubts have arisen:
- One of the approaches consists of the following:
Another popular approach is to build a slightly more complex model for predicting monetary value based on two sub-models: one for probability of response (e.g. using logistic regression as a function of RFM), and the other for revenue conditional on response (again, it could be as simple as a linear model of RFM). Expected monetary value is the product of the two predictions.
I already had my logistic regression output, but I don't know how to mix the response given by it (probability of response) in order to run, for instance, a linear regression for revenue. As you might realize, in the answer quoted above, it's said that it's done conditionally on response, but how can be this carried out? How can I mix those models?
Is it OK if I regress revenue (monetary) on the probability P given by logistic regression? I'm afraid this variable is somehow included since I obtained it by running a model with R, F and M as predictor variables.
Would you please provide me with some online resources (papers) to get more of this topic?
