Back in the days before big data and marketing automation, when the most effective communication channel was direct mail, marketers developed the technique known as RFM analysis.
RFM analysis is a simple way of segmenting existing customers based on readily available sales data. All you need to know is when a customer last bought something, how often they buy things, and how much they spend.
Fast forward to today, and RFM analysis is still a widely used technique. It can be used by itself; it can be the framework for a more granular segmentation strategy; it can be adapted to focus on things like engagement.
Whatever way you use it, RFM still requires those three pieces of data: recency, frequency, and monetary value.
Imagine you ran a grocery store. Over time, you’d get to know your customers: the regulars, the big spenders, the ones who drop in now and then to buy a pack of gum.
RFM analysis allows you to understand your customers in such a way, but on a vast scale. You do this by analyzing three data points:
When was the last time the customer bought something from your store? Or, in a digital marketing campaign, you may look at data points such most recent website visit, email click-through, or app login.
How often do customers come to the store? Again, you may choose to focus on each customer’s digital activity here. For a website, you might look at the number of unique sessions, number of pages visited or total logged-in time.
How much has the customer spent in the past? If you’re not looking at sales or e-commerce project, then you can use some other metric such as engagement, as long as you can assign a meaningful value to it.
Each one of these metrics is useful in itself, but they become a powerful analytic tool when you combine them.
For example, consider two customers who each placed $48,000 of orders in the past year. Will they respond to the same marketing messages?
Now, imagine that you perform some RFM analysis on these two customers. You discover that one customer has regularly spent $4,000 every month, while the other customer placed a $48,000 order at the start of the year and never contacted you again. Clearly, these two customers require very different marketing strategies.
There are plenty of ways to adapt this model, but let’s stick with the basic version. We’ll look recency, frequency, and monetary value of purchasing activity.
First, set a time frame for analysis, such as the last 12 months. Now, extract the following data from your sales reports:
You’ll end up with a table that looks something like this:
| Name | R | F | M |
| Customer 1 | 01/01 | 2 | $500 |
| Customer 2 | 03/04 | 10 | $100 |
| Customer 3 | 10/24 | 5 | $150 |
| Customer 4 | 06/06 | 8 | $50 |
What we’re interested in here are the relative values of these RFM figures. To that end, we can replace the data with a score from 1-4.
What we’re interested in here are the relative values of these RFM figures. To that end, we can replace the data with a score from 1-4.
We do this by sorting the table according to values and then sorting it into quartiles. For example, if sort by Recency, look at the 25 percent of most recent visitors, and give them all an R score of 4. Look at the next 25 percent and allocate them a score of 3, and so on.
If you perform that operation on the table above, you end up with something like this:
| Name | R | F | M |
| Customer 1 | 1 | 1 | 4 |
| Customer 2 | 2 | 4 | 2 |
| Customer 3 | 4 | 2 | 2 |
| Customer 4 | 3 | 3 | 1 |
Now you can see each customer’s profile at a glance. Customer 1 scores low on regularity and frequency, but high on monetary value. They may place another large order if you can entice them back.
Customer 4, on the other hand, is very regular and frequent, but they don’t spend a lot. They may not be seeing relevant offers, so that’s what we need to focus on with them.
The origins of RFM analysis are in direct mail and catalog sales. However, there are plenty of other applications. For example, in 2014, Delta Airlines used RFM analysis as the foundation for a new rewards scheme.
The issue with existing schemes, according to Delta, is that they reward miles traveled rather than dollars spent. A customer who regularly books discounted travel deals could potentially rack up a lot of free miles. On the other hand, a high-value customer who doesn’t travel very far (someone who books short trips at the last minute, for example) might miss out.
Offering rewards on a per-dollar basis helped encourage people to spend with Delta. Looking at recency (date of the customer’s last flight) and frequency (number of flight’s per year) allowed Delta to target the marketing for their reward program. Different messages were created for people who travel a lot versus those who take infrequent, long-distance flights.
Within the RFM model, each customer can have a score from 111 to 444, a total of 64 possible combinations.
That gives us sixty-four discrete customer personas based on just three pieces of basic sales data. Some of these personas are more important than others, however.
To identify the most important personas in your data, look for clusters. These are clumps of data that all sit together – for example, high-volume businesses will have a lot of people with a score of 111, due to their large amount of customer churn.
You can look for clusters by plotting individual customers on a scatter graph, and then looking for patterns. Whenever you see an unusually large section of your audience in one part of the graph, you’ve identified an important persona.
Some example of personas you might uncover include:
These are your best customers, and you want to keep it that way. If nurtured correctly, this segment can act as brand champions, bringing even more customers to your door.
Marketing messages: exclusive offers, pre-purchase of new products, premium customer support, refer-a-friend bonus
These people are great customers, but you can increase their lifetime value with the right offers. Find out what this segment needs and how to convince them to make you their supplier of choice.
Marketing messages: Loyalty schemes, volume discounts
A mismatch between recency and frequency shows that someone was a regular customer but has stopped using your service. This can indicate that something has gone wrong, whether their needs have changed, or they have had a bad customer experience.
Marketing messages: Satisfaction surveys, welcome back offers
People in this category have just discovered your business or possibly rediscovered you after some time away. Focus on relationship-building with this type of customer.
Marketing messages: Email list sign-ups, introductory offers,hints, tips, and useful content.
Some customers place a huge order and then disappear. In some cases, this may be because they came to you with a specific need and didn’t see any other offers that seemed relevant.
Marketing messages: Upgrade and maintenance offers, surveys
Low-scoring customers are the least promising prospects in your database. They don’t have a significant purchase history with your company and there have been no recent interactions. These customers will fall outside of the scope of most marketing campaigns. By focusing on more promising leads, you can invest resources where they’re likely to lead to results.
Marketing messages: Awareness-stage messaging, including content marketing and automated email.
With RFM analysis, as with any other type of analysis, you need to start by defining your objectives.
There are a number of things that you can achieve with RFM analysis, such as:
As we’ve seen, RFM analysis helps you to define some basic customer personas based on spending activity.
Hone these personas further by adding additional data. For example, your highest spenders might include distinct subgroups, such as commercial and retail customers.
Keep refining your personas until you have a solid grasp of your customer base. You’ll know you’ve achieved this when you start maxing out your conversion rates.
Customer personas help us to deliver the right message to the right person at the right time.
Even on the basis of RFM analysis, you can start to develop highly targeted marketing messages. After all, there’s no point in sending a loyalty offer to someone who never buys from you.
Similarly, there’s no point sending an introductory offer to someone who’s been a loyal customer for years.
Personalization is the key to moving prospects through a sales funnel, especially when you are using marketing automation tools. Know who you’re selling to, and know what they want to hear.
The purpose of Delta’s RTF analysis was to maximize the lifetime value (LTV) of each customer. LTV is the total amount that a customer spends with you, from the first purchase to the last.
To maximize LTV, you need to do two things. First, you need to encourage customers to spend as often as possible (which is reflected in Recency and Frequency).
Next, you have to encourage them to spend as much as possible in each visit (Monetary value.) If you can achieve both of these things, you’ll get the lion’s share of a customer’s LTV.
Recency and frequency are excellent indicators of clients who might be about to leave, or who have already left.
Sudden drops in these scores tell you that you need to take action – and fast. When you look at the monetary value of each client, you can get an idea of their potential LTV, as well as their purchasing habits.
Even if you can’t win back every client, doing this type of analysis can help you understand why some customers move on. Find out where you’re failing and take corrective action.
Where should you focus your marketing resources?
Probably not on customers with an RFM score of 111. This group typically has a low response rate and low return on marketing investment.
The same may be true of those with a score of 444. If a customer is close to their maximum LTV, you won’t see a significant ROI with them either.
RFM analysis can help you to identify the low-hanging fruit: customers who could increase their LTV with the right messaging. Focusing your efforts on the right groups will help you realize the best possible return on investment.
Is RFM analysis obsolete in a world of Big Data?
The examples above should demonstrate that no, RFM analysis is alive and well. It’s a simple and efficient way to use quantitative data – like sales records – to get a meaningful impression of clients.
Customer personas based on RFM analysis can easily be used in marketing automation. Take the insights from the RFM analysis, and use them to set up relevant campaigns. Your marketing automation should be able to get the right messages to the right customers based on their individual RFM scores.
But of course, you’ll get better results if you combine RFM analysis with more sophisticated analytics techniques. Work with a data specialist to find the essential customer insights that will power your marketing campaigns.
Finally, don’t forget the importance of marketing attribution. Only by tracking the outcomes of marketing efforts will you understand if your RFM analysis is giving you accurate information about your customers.