# RFM, Part 1

RFM. What does this stand for? My mother might mistake these letters for the name of one of my favorite college bands (REM). A consultant may nod his head confidently, too embarrassed to admit he is unfamiliar with the latest business acronym (CRM, ERP, BPR… they’re so hard to keep straight anymore). Ask somebody who has spent time analyzing data at a catalog retailer, and she will give the correct answer: recency, frequency, and monetary. These three simple variables can be used to determine the value of every customer in your database and can be used to develop a marketing plan to maximize future revenue streams from those customers. If you are analyzing customer data for the first time, building a simple RFM model is a great first step to segmenting your customer base.

In the first part of this three-part series, I will describe how to build an RFM model in a traditional retail environment, utilizing nothing more than your customer data and a simple Excel spreadsheet. The second part of the series, appearing in two weeks, will discuss how to apply an RFM model to an Internet media company. The third part of the series, appearing in four weeks, will discuss how to use the results of the RFM analysis to develop successful marketing strategies targeting each segment of the customer base. So, without any further delay…

Defining the Scoring System

Customers in the database will be scored on each of the three variables mentioned above: recency, frequency, and monetary. The score will be either a 1, 2, or 3. A score of 1 indicates that the customer ranks low within that specific variable. A score of 3 indicates that the customer ranks high within that specific variable. Let’s look at what each of the variables means:

• Recency is the time that has elapsed since the customer made his most recent purchase. A customer who made his most recent purchase last month will receive a higher recency score than a customer who made his most recent purchase three years ago.
• Frequency is the total number of purchases that a customer has made within a designated period of time. A customer who made six purchases in the last three years would receive a higher frequency score than a customer who made one purchase in the last three years.
• Monetary is each customer’s average purchase amount. A customer who averages a \$100 purchase amount would receive a higher monetary score than a customer who averages a \$20 purchase amount (average purchase amount = total dollars spent on purchases in last three years / total number of purchases in last three years).

The question that may have arisen in your mind is how you determine who gets what score for each variable. The simple answer is that it depends on your business model. A low-volume, high-purchase-amount business will define the parameters of the scoring system much differently than a high-volume, low-purchase-amount business. Illustratively, a car dealer may define the parameters of the scoring system as follows:

Recency
1 = Customers who made a purchase more than 48 months ago
2 = Customers who made a purchase more than 12 months ago but fewer than 48 months ago
3 = Customers who made a purchase in the last 12 months

Frequency
1 = Customers who made a single purchase in the past 60 months
2 = Customers who made two purchases in the past 60 months
3 = Customers who made three or more purchases in the past 60 months

Monetary
1 = Customers with an average purchase amount up to \$20,000
2 = Customers with an average purchase amount from \$20,001 to \$40,000
3 = Customers with an average purchase amount greater than \$40,000

Therefore, a customer receiving a score of 3 in each variable (333) made a purchase in the last 12 months, made three or more purchases in the past 60 months, and has an average purchase amount greater than \$40,000. Obviously, this customer is very valuable to the car dealer. Conversely, a customer scoring a 1 in each variable (111) is less valuable to the car dealer.

The corner pharmacy may define the parameters of the scoring system as follows:

Recency
1 = Customers who made a purchase more than nine months ago
2 = Customers who made a purchase more than three months ago but fewer than nine months ago
3 = Customers who made a purchase in the last three months

Frequency
1 = Customers who made a purchase in the past 12 months
2 = Customers who made 2 to 12 purchases in the past 12 months
3 = Customers who made 13 or more purchases in the past 12 months

Monetary
1 = Customers with an average purchase amount up to \$15
2 = Customers with an average purchase amount from \$16 to \$45
3 = Customers with an average purchase amount greater than \$45

The point here is that every business is different and, therefore, you must define the parameters of the scoring system based upon your unique business model.

Building the Model

Now let’s get started building your RFM model. Pull a list of every customer who has made a purchase during a reasonable purchase cycle. In the car dealer example, I assumed a reasonable purchase cycle of 60 months, or five years. The logic behind that number is that if a customer hasn’t purchased a second car from the dealer in five years, it is likely the customer bought a car someplace else and isn’t coming back to the original car dealer anytime soon. Use similar logic to determine a reasonable purchase cycle for your business. Make sure you pull the following data for each customer record: date of most recent purchase, total purchases during the purchase cycle, and average purchase amount during the purchase cycle.

First, score the customer database for recency. Sort the customer base by date of most recent purchase, with the most recent purchase date at the top descending to least recent purchase date at the bottom. Assign a score of 3 to the top 20 percent of customers, a score of 2 to the middle 60 percent of customers, and a score of 1 to the bottom 20 percent of the customers. Your recency segmentation is complete. Relatively easy, isn’t it? Repeat the process above for the frequency and monetary variables so that every customer is assigned a score of 1, 2, or 3 for each of the three variables.

Once completed, sort the customer list so that the customers receiving scores of 333 appear together to form a single segment, the customers receiving scores of 323 appear together to form a single segment, the customers receiving scores of 322 appear together to form a single segment, and so on (remember that the numbers are always represented in RFM order). When this is complete, you will have 27 total customer segments, one segment for each combination of the RFM variables (listed below). Identical to the car dealer example above, customers who score 333 are your most valuable customers. Customers who score 111 are your least valuable customers.

 111 222 333 121 212 323 131 232 313 112 211 332 113 213 331 122 231 321 123 233 322 132 221 312 133 223 311

How can this model be applied to an Internet media company? We’ll cover that in two weeks — same bat time, same bat channel.