Press enter to see results or esc to cancel.

RFM Model
Elena Dobre

Step by Step Guide To Building An RFM Model (and How to Use It)

There was a time when advertising was everything.

People would spend huge budgets on getting attention towards their brand and acquire new customers. Over and over again.

Until terms like “retention”, “loyalty” and “building a relation” became popular.

We find ourselves in an era where the relation between consumers and providers is cold. There’s no emotion or commitment between the two parts. As a marketer, I believe that we need to shift the focus from being orientated on transaction to putting the customer in the first place.

I’ve seen many online businesses focusing on reducing their Customer Acquisition Cost (CAC). They completely disregard retention. This is similar to doing cold-calling and expecting everyone to buy the first time they encounter your product.

To me, that just sounds like bad focus.

Why spending 5 to 10 times more on acquiring new customers rather than keeping the ones you already have? The reality is that companies with 24 months of activity in online get 50% of their revenue from returning visitors.

RFM Model Graph

Screenshot from RJMetrics Ecommerce Growth Report 2015

It’s ineffective to focus on acquisition rather than retention.

Let me give you an example…

I’ve worked with a company that used to spend 20 times more on getting new customers. They had huge advertising budgets. Since the focus was on acquisition, they tried their best to reduce CPMs and CPAs, but they didn’t do anything about the Customer Lifetime Value.

RFM Model Intro

Source

If you are in a similar situation, this step by step guide will help you shift the focus from acquisition to retention. I’ll show you how to build an RFM Model and how to use it to build a strong relation with your current customers.  

Let’s start!

Step 1. Tracking

First of all, have in mind that when you try something new for your business, you need to find a logic. I recommend starting with building a database containing the history of the purchases from the past 3 years.

From a first look at the data, you will observe some patterns. When I once analyzed a customer database, I observed that there were customers who bought once but never purchased again from the website. On the other hand, we had customers that purchased every time we launched a new product.

The first major insight was that these customers had different habits and we needed to approach them differently.

When you have the database, you need to select some criteria that will help segment the customer database and give scores. The most important metrics that you should consider are:

  • Revenue (1)
  • Number of orders (2)
  • Average Order Value (3)
  • Last Order Date (4)
  • Last order in days (5)

RFM Model Segmentation

Right after deciding what you need to track, you have to flag all of the previous purchases according to:

  • Recency – the time when they last placed an order;
  • Frequency – how many orders they have placed in the given period;
  • Monetary value – how many dollars they spent since their first purchase (Customer Lifetime Value)

Step 2. Segmentation

The second step in the building the RFM Model for your eCommerce site is defining the categories of customers. You will have to define them based on recency, frequency and monetary value.

Recency

If you decide to make the analysis for the last three years, divide the time interval into 4 or 5 parts. List all the purchases from the most recent to the least recent and divide the time equally. You will get five segments:

  1. Old
  2. Lost
  3. Potential
  4. Regular
  5. New

RFM Model Criteria

Then, assign numbers to each customer: 1,2,3,4,5.

This is an example with four-time intervals. That’s why you see only max. 4 points assigned to Recency.

In this phase, you should be able to have a major insight about your customers. For instance, I once discovered that users who bought more recently were more inclined to open our newsletters and visit the website. They were more engaged.

“Users who made recent purchases were more willing to visit from email and interact with the website.”

You have to repeat the process for Frequency and Monetary Value as well.

Frequency

Sort the entire file containing the purchases from the highest number of purchases to the lowest number of purchases. Then, divide it into equal parts like you did with the recency.

For example…

If you have divided the time interval into 5 parts, assign points from 1 to 5 for the frequency as well.

Pro tip: Create customer surveys triggered to find out how much they are willing to spend on your site.
 

Here are some questions to ask in customer surveys:

  • How much are you willing to spend for [product X]/ [Special event], etc.?
  • What would you like to improve in the products on [your site]?

When you analyze the answers, pay attention to the differences between the users who make recurring purchases and the ones who purchase only once. If the difference is significant, then frequency is important to your business.

In my case, when I analyzed the results, the difference between the two was not as significant as in the Recency research. Therefore, it was more important for that business to have recent purchases rather than having a good frequency score.

Monetary Value

The third factor in the customer database segmentation using the RFM Model is the monetary value. The monetary value is the overall amount of dollars that your customers had spent in their lifetime.

All you have to do is sorting the list from the highest amount of dollars spent on your site to the lowest and divide the database into 4 or 5 equal parts (depending on how you decided to do it). Then attribute points to each part.

After applying these three segmentation criteria you’ll get a list of three-digit codes, which represents the RFM score for each customer.

Step 3. Testing

Now that you have segmented the customer database based on recency, frequency, and monetary value, how do you know which segments are the most profitable?

You test them!

First, you have to extract a sample size from the total number of customers (segmented as discussed below) and start a campaign. Choose a sample size of 5% or 10% of the total number of customers included in your database.


170,000 clients, divided by 8,500 (5% of total) equals 20. This means you have to pick every 20th client in the database.


Use this rule if you want to have a sample that is representative for the entire list.

By testing 5% or 10% of your list, you minimize the risks of launching a promotion campaign that would fail. If you need help with setting up A/B testing experiments and personalization workflows for your segments, Omniconvert is the right technology to do it. I know I’m biased, but it’s the only way to do it without installing multiple codes on your site.

Step 4. Analysis

To analyze the results of your campaign, you should measure the Conversion Rate of each segment that has an RFM score attributed. The segments with the highest conversion rate are the ones that will get you the highest amount of money to your business.

When you analyze the results, use this simple procedure that works anytime, for anyone. You don’t need sophisticated tools for this.

First, you need to divide the total costs of the campaign with the total revenue.


If Revenue/Cost is 1 or greater than 1, then it’s profitable to keep investing in that customer.


Draw a line below the row that contains the last value greater than 1. If it’s below one, it means that it’s ineffective to invest in that type of customer or that you should run more tests to find out what’s making him behave like the ideal (read most profitable) customer. I call this method “waking up the customers that sleep.”

P.S.: We have created an RFM template to help you visualize the whole story better. Take a look.

Step 5. Iterating and growing

If you build the RFM model the right way, you will be more persuasive in your communication with your customers. It will most likely convince them to continue as your client. Check out these ideas to increase sales if you feel like you don’t know where to start from.

For example, if your customers are in the “1” category on Recency, you can send them a re-engaging newsletter. This newsletter could mention all the new offers you have on your site, any changes in design, or how the overall experience has been improved since they last visited your site.

If you want to talk to customers that are in the “5” segment of your Monetary Value, let them know how much they mean to you and your business. Give them special discounts if you can, tell them they’re VIPs, make them know they are appreciated. This will only bring them back to your site and buy more.

Finally, I encourage you to be creative. You can use the RFM model in any way suits your business. But remember that just doing the RFM analysis and segmentation will not influence your website’s results. You will have to act based on the insights and start testing approaches and targeting them in a more emotion-based way. In the end, you’ll get more conversions.

rfm model cta

Comments

4 Comments

John

Hi Andra. Just a head’s up that this is not the correct construct for the RFM model as originated by Harte-Hanks. While your model may rely on similar variables, it isn’t true RFM in the perfect sense of the definition. One hint: for each different element you should always end up with the same number of clients in each bucket. 5’s should be the same as 4’s, etc. Rather than identifying specific cut-offs for recent purchases, you allow the data to reveal the patterns. Likewise, calculating F and M is also more complex than what is describe. Not trying to be a troll. And certainly some method for segmentation is better than none. But wanted to give a head’s up that RFM in its purest form is quite different…

Elena Dobre

Hi John,

Elena here, in charge of Marketizator’s blog management.
Thank you for pointing this out. I will make sure to update this article as soon as possible.

Best regards,
Elena.

Amin Jamri

Hye Elena

Can i have the upgrade article on the RFM Model since last article was given a comment.

Thank You


Leave a Comment

CommentLuv badge

165 Shares
Share74
Tweet24
Share67