Calculating Customer Lifetime Value (CLV) with Recency, Frequency, and Monetary (RFM)
CLV, "the present value of future cash flow attributed to a customer's tenure with a company", can be determined by variety of approaches. Based on a simple equation, one can calculate the cumulative profit (value) from a customer based on assumptions such as retention rate and historical profit margin from customers. However, a customer's retention can be influenced by factors such as demographics (age, geography, education background), behavior (Recency, Purchase Frequency, Monetary Contribution), competitions, peer influences, etc.. As a result, a dynamic approach should be taken to account for such variations.
In this repo, I calculated CLV by predicting the retention rate (r) of customers' future purchasing cycle using Logistic Regression based on his/her Recency of purchase, purchase Frequency, and Monetary contribution from past purchases.
RFM
By definition, RFM represents:
- R(ecency): how recently did customer purchase?
- F(rquency): how often do customer purchase?
- M(onetary Value): how much do they spend (each time on average)?
The determination of Recency, Frequency, and Monetary of a customer, should provide insights to the following:
- How to segment customers to determine who are more likely to response to ads / to purchase?
- Which type of customers to send ads in order to breakeven and make profit?
Data
The CDNow data set (download at here), consists of 69,659 transaction records by 23,570 unique customers, captures the purchase records between Jan 1997 and June 1998.
Code
- Source code (R)
References
This project was largely based on the work done by Jack Han at Data Apple. However, new libraries/functions are used to update data transformation and visualization of data such as dplyr and ggviz.
Other references used:
created by stackedit.io