What is the Recency Frequency Monetary (RFM) customer segmentation model?

Please go to my personal website if you want to know more about me or get in touch.

I want to segment my customers

Build your customer segmentation easily just by uploading a CSV with our self-service analytics platform

Create your free account

In this guide, you will understand what is the RFM (recency frequency monetary) behavioural customer segmentation and why the RFM segmentation can help you develop your business. Everything you need about general customer segmentation can be found here.

Definition of the marketing RFM segmentation

The RFM segmentation is maybe one of the most popular and long-time used customer segmentation technique by marketers from all-sized companies. The concept is quite straight forward and it helps many companies throughout the world understanding better about their customer behaviour.

Recency - the "R" of RFM.

Recency dimension assumes that the more recent you've been purchasing from the company, the better. Customer will definitely remember you and have your company name top of mind. They will still be receptive to your communication and still might want to interact with your companies, let's say, to get feedbacks and support in using your product or service. At the other side of the scale, customer who purchased 360 days ago might already be considered at churners. However, this might not be true based on the lifecycle span of your product or service.

Let's say you're selling high quality design bed sheets. Chances are that your customer will buy afford them and think about purchasing new ones only every 3-4 years. No hearing from a customer in the meantime doesn't mean he totally forgot about you or just don't want to have anything to do with you. It's just that the sales lifecycle for this product category is longer than for any other staple.

Frequency - the "F" of RFM

The question is this time not about time, but on how often. How many transactions / purchases a customer made in the timespan considered. Did he buy only once? Did we buy 12 times and went 12 times through the checkout process? This actually makes a big difference. Imagine you would have buy 12 times a year - in a brick and mortar business environement, you would expect sales people to recognize you whenever you come in, get exclusive access to new products, be invited to some premier. Virtuality doesn't stop this need for acknowledgement. 

Monetary - the "M" of RFM

Here comes the most obvious and straightforward dimension: money. More precisely, the amount of money spent over the time span. Monetary must be understood using the F dimension: what the amount spent in 1 single purchase or accross many purchases? In case you have multiple currencies to handle, you should convert all the money in one single currency with an even foreign exchange rate. Just saying. 

Upload a CSV now and score your customer using the RFM (Recency, Frequency, Monetary) methodology

Create your free account

Why you should use a customer RFM segmentation?

We are living in a world of constraint: constraint of money, limited budget, limited time. RFM helps segmenting a customer database based on their past behaviours and allows marketers to develop fairly accurate and targeted strategies where this is relevant. Acquiring new customers is actually very expensive for a company: marketing money needs to be spent on search engine acquisition, discounts and free trials must be offered, pricing proposition must be drafted to get the contract and beat the competition. Fidelization however is much cheaper: the customer know you, you know him, he already purchased at your shop and got a tast of what's like to be a customer of yours. You also captured information about who they are, their contact details such as email or phone number, making it possible for you to reach to them directly. 

Two types of customer segmentation: a segmentation based on who the customer is and on how he behaves. RFM segmentation is based on the behaviour of the customer toward your shop. How often do they purchase? How much do they spend? When was their latest purchase at your shop? This behavioural customer segmentation allows to make decision based on the observed attitude of your customers. No bias is being made due to their age or average income for instance. Sometimes, acts are better than declaration. 

RFM segmentation is relative. And this is quite interesting.

When you perform a RFM customer segmentation exercise, it's not about absolute but relatives. Customers get points attributed based on their relative location in the scale of each dimension. To use a word coming from the statistical world, we are talking about quintiles. All customers are scored relatively against one another. In other words, the relative quintile scoring approach allows to factor in wheter your customers purchase on average every 2 years or once per month (lifecycle) - or if average basket size is $30 or $4,500. It just doesn't matter anymore - and that's the beauty of it.

This 'relative' approach allows to factor in the specificities of each business. Not every business looks like the same. If your activity is about selling curtains, then chances are that your customer will visit you every 4 or 5 years only, and that will be ok. If you sell toilet paper, your customers might have to order very often at your place. Same phenomenom happens in terms of basket size, depending on the vertical your are into. If you operate in multiple countries, a "good" customer in country A might be different from a "good" customer in country B, depending on your current penetration and market share in each market. In this end, this is up to you: one can decide to run one single RFM segmentation for all the customers wherever they are, or multiple RFM segmentations for each market. It all comes down to your marketing strategy.

Customer profiles you will get with an RFM segmentation

Customers who purchased only once but spent a huge amount, 6 months ago: what would you do with them? Maybe try to understand what was the purchase they made in first place. If this their basket was high because they bought a very expensive item, they might not be likely to purchase a similar item soon. However, if their basket was a collection of many items, this is good and you might want them to purchase ofter at your shop.

Customers who purchased many times a year but every time small amounts... maybe you should try to upsell them and increase their basket size? One idea could be to personnalize their experience and push the cross-sell offers. You can also build a targeted newsletter and promote other product categories they are not aware of.

Customer who purchased recently with an average ticket, but already made 2 purchases... you need to build the relationship with them at the first transactions matter in a business relationship. They start to be engaged customers. You might want to retain them. 

The RFM model's limitations and warnings

Prediction are limited - since the model is mainly descriptive it doesn't tell you directly what to do next in terms of sales and marketing actions. To build a more prediction oriented model, RFM would need to be used along with other inputs and techniques. Predictive analytics are techniques involving the usage of machine learning, regression models (linear or logistic) with some proability behind. The downside of these techniques is that it might appear to the end-user as a black box - and marketers usually are afraid of too much black box when they need to spend their budget on them. 

All the customer segment need to be considered when implementing marketing actions. The danger of the RFM customer segmentation technique is that it can give the wrong impression that some segments are not worth investing. Which is wrong: the investment of today makes the profit of tomorrow; hence neglecting a customer segmentation because "it's not worth marketing money to spend on it" is definitely a wrong strategy. Don't under invest segments, just because "they are not good". The purpose of the RFM segmentation is to get an customized marketing approach for groups of customers having the same profile, not to cherry pick between segments!

Run RFM segmentation every month of every quarter at least. Your database evolves - and your customer knowledge should always be up to date. You would need to run RFM analysis as often as required to keep track on the latest evolution of your customers. If you store the results (e.g. in an excel spreadsheet), you would also be able to compare for each customers the evolution of their RFM scoring over time. This could also be a good indicator is you manage to create momentum and get traction following your marketing actions.

How to interpret the RFM scoring model's output? 

The classical RFM model approach is to use for each of the 3 dimensions which are RecencyFrequencyModel, about 5 notes. 1 being the lowest score a customer can achieve on (lower-end), while 5 points will be attributed to the best in class customer. So far so good. Having 5 degrees allow to real fine-tune the outcome and understand truely the consumer's behaviour.  If we do some maths together, that would be that we have 5^3 combinations possible. That's 125 theoretical combinations possible. 3 axes (or dimensions): that's typically a cube kind of visualization. 

Building a 3D cube to visualize the output is definitely not the way to go. Too complex and while computer can do that brillantly, our cognitively limited human brains won't find it easy to interprete. Let alone for business people who are likely to use Excel as a data companion tool. In terms of visualization, to make it understandable by everyone, the trick is the following:

  • Bundle R (recency) and F (frequency) scores together and create the segment from here. This gives already quite enough information about the usage pattern and interaction.
  • Get the M (monetary) as a separate layer, coming on top of it. It's an additional information.

You would have for instance a segment Potential loyalist:

  • Potential loyalist - low $
  • Potential loyalist - middle $
  • Potential loyalist - high $

In terms of presentation, while the raw data is available for the 125 segments, some of the data has been regrouped - Eg. monetary has 3 final level, while the scoring is available for 5 (each quintile). It makes it easier for marketers to read and interpret. Remember: a model is useful only if marketers can make something out of it.

Interested? Want to know more on what RFM means for your customer database? Drop us a note!

This post was rated: 5 / 5 by 1 readers
You rated: Not rated yet... but please be kind!

I want to segment my customers

We develop self-service analytics solution allowing to build a customer scoring based on their behaviours.

Create your free account