This paper discusses the basic postulates of Recency, Frequency & Monetary (RFM) analysis, a heuristic modeling approach, used in Predictive Analytics, to segment a target market into preferred segments and not so preferred segments. The preferred segments are characterized by their high response rate or high willingness to purchase as opposed to other segments which are not as preferred.
The paper also exemplifies these concepts with the help of a case study in the tele- communication sector where a company uses an existing data base to arrive at RFM categorization as well as identifies the profile of customers in the preferred segments.
Introduction
Predictive modeling, the way it is understood in the Business Analytics context, is a way of predicting consumer behavior by analyzing a database either existing in the company concerned or on a database created with the help of an empirical survey. Essentially, a modeling approach, predictive modeling helps the company to identify profiles of consumers who would be more likely to purchase a product or a service which the company might be offering to a specified and defined target market. Applications of predictive modeling can be seen over different industries and in different managerial functions.
For instance, for an entrepreneur offering a new product in a specified target market, predictive modeling can help in understanding the consumer needs and preferences with respect to the attributes defining the product. For a service oriented company it can help to determine the profile of the most preferred segment and predict the percentage of customers who may actually purchase a new service being offered. For a credit card company or an organization offering loans of any kind, predictive modeling may evolve guidelines as to what kind of consumer profile would merit a preferred treatment and to whom loans may be extended with a softer level of interest.
Once the predictive modeling context is well understood and the objective in terms of what phenomenon is to be predicted has been clearly stated, the approach would define measurable variables for each item of the in the situation … the predictor variables… as well as the variable to be predicted … the dependent or target variable.
Thereafter the predictive modeling uses either:
(a)Analytical approach like Logistics regression, Linear Regression Analysis, Factor and Cluster Analysis, Conjoint analysis, or
(b)Heuristic approach… like RFM analysis, or
(c) Data mining approach, which combines Heuristic and Statistical approach such as Classification Trees.
This paper proposes to discuss the basic concepts of the Heuristic Approach of RFM Analysis and provide an example of RFM Analysis applied on the database of a company operating in the telecommunication field in India.
Predictive Modeling and RFM analysis.
In strategic decision making companies often strive to determine who are the most valuable customers whom they would give special privileges to, invest to build up long term relations with, say in a CRM scenario, or target offers for mail orders, catalogue buying or any kind of direct marketing initiatives. The objective, in most of such situations, is to find out who the most likely buyers are, who makes purchases most frequently, who spend the most and who have the greater probability of coming back for repurchase. In many such initiatives, RFM analysis, recency-frequency-monetary analysis, helps identify consumer segments and customer profiles having such characteristics.
‘The fundamental premise underlying RFM analysis is that customers who have purchased recently – , have made more purchases and have made larger purchases are more likely to respond to your offering than other customers who have purchased less recently, less often and in smaller amounts.’
[Charlotte Mason, 2003, University of North Carolina].
The analysis helps an organization to focus on a smaller section of the target population which again follows another managerial premise, Pareto Principle that 80 % of the business comes from 20 % of the customers.
In the past 30 years, direct mailing marketers for non-profit organizations have used an informal RFM analysis to target their mailings to customers most likely to make donations. The reasoning behind RFM was simple: people who donated once were more likely to donate again. Currently, with the availability of CRM software and the use of e-mail marketing, RFM analysis has become an even more important tool. Using RFM analysis, customers are assigned a ranking number of 1,2,3,4, or 5 (with 5 being highest) for each RFM parameter. The three scores together are referred to as an RFM composite score. The database is sorted to determine which customers have been the best customers in the past, with a composite score “111″ being ideal. Of course, in some organizations marketers consider 5 to be the most preferred RFM parameter, in which case ‘555’ would be the most preferred customer.
(http://searchdatamanagement.techtarget.com/sDefinition/0,290660,sid91_gci751219,00.html)
There are many justifications as to why RFM analysis works. Customers who bought most recently from an organization, are more likely to respond to the next promotion than those whose last purchase has been way back in the past. This is a universal marketing phenomenon and has been observed in many industries such as insurance, banks, cataloging, retail, travel, etc. In a similar manner, customers who have purchased frequently are more likely to respond than the less frequent ones. Also customers who are big spenders often exhibit much higher response rates than small spenders.
HOW CAN ANALYTICS HELP
Analytics is increasingly being used as a tool to solve complex organizational problems – leading to better decisions. These are the decisions which were once taken solely by gut instincts.
The success of any company in the Telecom industry currently depends on two broad factors –
- Ability to add new subscribers (both data and voice)
- Ability to retain existing subscribers (Since, Mobile Number Portability is now available by all operators)
This paper focuses on the second part, which is on the indicators which would help the company minimize the tendency of subscribers to switch from their service to others. One of the major indicators of this tendency is measured by Port-in Port-out ratio (it is also commonly referred to as Churn).
The number of subscribers switching to a given provider from others is referred to as Port-in. Port-out indicates the number of subscribers switching to a different provider from the given company. A port-in port-out ratio of less than 1, hence, is good for the company because more subscribers are coming in than out. If the ratio is greater than 1, it is considered bad for the company.
There are two ways of making this ratio healthy – increase the number of port-ins or decrease the number of port-outs. In order to do so, the company would need to strategically connect with the individual subscriber base – the better their needs are taken care of, the less likely it is that they will switch to other provider.
Keeping the above fact in mind, a telecom provider usually comes up with a number of plans to woo the existing customers. This comes with a catch though. It is almost impossible to roll out tailor made plans for every subscriber – such would be too expensive. At the same time, covering the entire subscriber base with a few plans would not go down well with specific consumers whose needs might be different.
One solution might be to come up with a plan, say, ‘Pay-Per-Use’ plan – and in order to do so, a broad survey of consumers is required – most of their usage details are already with the company. It is their preferences which need to be mapped with their eagerness to take up a new plan from the same provider instead of switching to another service provider. This new service was called ‘dataplan’ by the telecommunication company.
Thereafter the company wished to use Analytics to help identify, who among the existing subscribers are most willing to take up the new plans. It is with this end in mind, that the data has been collected, data base was constructed and analyzed. The database elements have been discussed in the next section.
Authors
Raghuveer Kodali holds a Bachelor’s degree in Electrical and Electronic engineering, from DVR & DR and is currently pursuing his MBA from Myra School of Business, Mysore. HS MIC College of Technology. His Internship in Market Research for a product UNIled In Kwality Photonics. He is Interested in Analytics and Market Research.-
Nitish holds an MBA degree in Marketing & Strategy from MYRA School of Business, Mysore. He has done his Bachelor of Technology in Mechanical Engineering and has about two years of experience in manufacturing domain. He is an automobile enthusiast and likes to travel. -
Sangitha Ajith is an MBA student in Marketing & Strategy at MYRA School of Business. She holds a degree in Bachelors of Commerce from Mysore University. She spent her summer interning in Brandcomm as marketing analyst. Her interests lie in creative art. 
Sri Valli holds a PGDM degree in Finance & Analytics from MYRA School of Business, Mysore. She had done Bachelor of Technology in Computer Science and Engineering. She is an avid puzzle solver with a special interest in Rubik’s cube.
Ashutosh Kar holds a Bachelors in Computer Science & Engineering from Jaypee University of Information Technology. After working with Data Warehouses and Business Intelligence tools for 6 years in various IT companies like ORACLE & IBM, his interests shifted to finance and mathematics. He is currently pursuing his MBA from MYRA School of Business, Mysore. His academic interests include dynamic optimization and optimal control theory.
Mihir Ghosh: Academic Background – PGPX Executive MBA from MYRA School of Business. Graduated from IIT Kharagpur in the field of Dairy and Food Engineering. Experience – 7 Yrs and 4 months professional experience in project sales and marketing in food, pharmaceutical and life science sector.
