How to Calculate and Predict the Value of Your Customers — RFM

What is the ultimate metric to calculate the total value of a company?

In order to find the right value, a company should be divided into its smallest unit: one customer. The sum of each customer’s equity makes the total (current) value of the company. It all starts with RFM, so this approach will be our starting point.

This article series will be of two parts and this is the first part where I will be discussing about the importance of RFM as well as its metrics and scores.

Customer equity can be calculated using three metrics: Recency, Frequency and Monetary. We are living in a world where every action is transformed into data such as purchase history, browsing history, prior campaign response etc. So it’s no big deal for a company to reach the data and use these metrics in order to calculate RFM of each customer. RFM, alias for Recency, Frequency and Monetary is a rule-based methodology that uses some metrics to determine quantitatively the metrics and the score of each customer in order to segment them into various categories based on their purchase habits. Let’s take a closer look at these three metrics:

Recency shows how many days ago did the customer last purchase,

Frequency shows how often the customer purchases in a specified time period,

Monetary shows how much money did the customer spend in a specified time period .

RFM numerically ranks each customer with 5 for the highest frequency, highest monetary and the lowest recency. A customer who purchases more frequently and who spends the most has 5 for the F and M score whereas for R score, this is the opposite, because the customer who had just purchased -meaning the customer with a lower recency value- is more valuable and therefore deserves a 5 score for recency.

With the RFM scores each and every customer can be categorized and finally it would be easy to aggregate tens thousands of customers into ten segments. Having an effective CRM with ten segments would be easy, right? It also makes it possible to see each customer’s current behaviour to the company, whether he/she is a Champion, a Loyal Customer, a customer that Needs Attention or a customer that is About to Sleep. In a non-contractual setting, it’s vital to detect a customer who is thinking of leaving the company because there is no end date and the customer wouldn’t notify you before he/she leaves. With the aforementioned RFM scores, it’s easy to detect them and develop the best strategy in order to make them stay and purchase. I recently shared a Kaggle notebook where I implemented this approach. The link to my notebook is here.

Recency is the most important metric in RFM. It is the key to detect who is “alive”. Being alive is a metaphor for “a customer who still purchases from the company”. But being alive is a term that should be considered differently for every customer due to their different frequency values. Imagine there is a customer that used to purchase once in every week and didn’t purchase for four weeks, it’s probably because this customer is not alive. But for another customer who purchases once in every two months, not purchasing for four weeks is not a sign of being no longer alive.

RFM is the key to see the current position of the customers with metrics and scores for each and every customer but it’s also very important for a company to have a projection of what’s coming next. There is no proof that a customer who is scored as a Champion would continue on their purchasing habits for years. It shouldn’t be forgotten that there is always a risk for churn. It’s not possible to calculate the overall value of the company with RFM scores. Moreover segmenting a new customer that has only purchased for once is not possible with RFM. These problems will be tackled in the second part.

The first part of this series was about “today”. In the second part of this series which will cover the “future”, I will discuss about the customer’s life time value, CLTV and two different approaches that is needed to calculate this value: BG/NBD and Gamma Gamma Models.


  1. Data Science and Machine Learning Bootcamp Lectures