Is The RFM Model Obsolete?


The Recency, Frequency and Monetary Value model has provided practical applications over the years, particularly in direct marketing. With the advent of new-age tech, there are apprehensions that the model no longer serves its purpose.

You may not know Italian economist Vilfredo Pareto, but chances are that you are familiar with the Pareto Principle or the 80/20 rule. It states that 80% of the consequences come from 20% of the causes. In business terms – 20% of a company’s customers account for 80% of its overall revenue. 

This makes it critical to have a clear understanding of who these high-value customers are. This is where customer segmentation comes in – the process of grouping consumers based on shared behavior or other similar characteristics. The main objective is to identify a high-value clientele for direct marketing, which further helps design overall marketing strategy and planning.

Arthur Hughes popularized the RFM (Recency, Frequency and Monetary) Model which considers the time period of the last transaction, the frequency with which the customer transacts with the brand, and the total monetary value a customer spends. Although it has been in use to develop marketing strategies for many years now, several pitfalls exist while using the RFM technique.

Losing its relevance

With the emergence of new-age tech, including artificial intelligence, machine learning, and neural networks across businesses, the model seems outdated. Let’s have a look at some of the pressing reasons backing this hypothesis:

Narrow scope: RFM solely concentrates on the best consumers in order to identify valuable clients in businesses. When the majority of customers do not buy frequently, spend little, or have recently made a purchase, it offers nothing in the way of a significant score on recency, frequency, and monetary factors. Additionally, RFM estimates a single response model for every customer in the database, and hence makes the assumption that the customer database is homogenous, which most often is not the case. 

Insufficient amalgamation of factors affecting customers: Other than recency, frequency, and monetary value, a customer’s propensity to respond to marketing stimuli depends on a range of elements which is not captured by the model, which includes;

  • Demographic data such as age, sex, gender, ethnicity, employment, etc
  • Behavioral data, including brand preferences, life events,
  • Psychographic data such as beliefs, personality, priorities, and even
  • Geographical data

Moreover, the RFM models are based on the customer perspective and do not consider the product perspective

As a result, the RFM model fails to provide a complete picture of customer behavior. On the flip side, big brands involving Disney, IKEA, and JetBlue rely heavily on demographic and psychographic data to provide a better and personalized CX. 

Limited predictive modeling: The RFM model relies heavily on just the three elements of historical data of the customers; hence the system lacks precise prediction of customer behavior. It tells marketers nothing about how the customer will respond to a particular campaign. The model builds an assumption that the brand’s best buyers will be their responders. This ignores the fact that customer behavior might change over a while or have already changed due to many other factors.

Failing to answer the why: The model correctly points out the most engaged or valuable customer (answers ‘who’) based on data; however, it fails to answer a simple question as to why these customers remain committed to a brand. For example, a brand wishes to expand its business to a new geographical location. It wants to enjoy the same level of engagement with customers as in its current stores. Is it possible to replicate the success of one region in another without even having a fair understanding of why the customers are there in the first place? Short answer, no. 

Inability to act on real-time data: Ecommerce businesses are evolving and growing every day. In the world of data analytics and advertising, instant campaigns and tactics are a must. As a result, the machine learning algorithm’s self-learning and automatic segment analysis are also more valuable and practical. It’s high time for businesses to incorporate platforms feeding real-time data and gaining valuable insights as output.

Finally, the RFM model cannot be applied to new customers as it only predicts the present set of customers.   

The way forward

Machine learning algorithms’ self-learning and autonomous segment analysis are seen as more viable and practical solutions. However, given the shortcomings of RFM models, efforts can be made to strengthen their predictability by developing new models or including additional variables to forecast the behavior of customers.