Some years ago, I created a software product with the motto:
“Every customer counts.”
It sounded right. It sounded customer-centric. It even felt like a principle worth building a company around.
It took me years to realise that it was wrong.
Every customer deserves respect, but not every customer contributes equally to a business.
Some customers buy frequently, trust you, recommend you and help you improve. Some require far more time and support than the value they create. Some are simply there for the free resources and, however long you nurture them, may never engage commercially. Others become a burden. They drain your team’s energy, consume resources and introduce complexity without creating a healthy relationship in return.
Then there are the customers who look small today but share many of the characteristics of your best customers. With the right attention, they may become your most valuable relationships tomorrow.
The problem is that most businesses treat all these people in roughly the same way.
They place them in the same CRM. Send them the same newsletter. Offer them the same discounts. Give them the same level of attention. Sometimes, they even spend more time trying to satisfy the most demanding and least valuable customers than protecting the customers who keep the business alive.
That is why RFM matters.
It provides a structured way to stop treating your customer database as one homogeneous audience. It helps you understand who is creating value, who could create value, who is slipping away and who is unlikely to become a meaningful customer regardless of how many campaigns you send.
“Every customer counts” was a good intention.
A better principle would be:
Every customer should be understood. But not every customer should be treated the same.
RFM stands for:
The principle behind the model is simple.
A customer who bought recently is generally more likely to buy again than someone who has been inactive for two years.
A customer who has purchased ten times has demonstrated a stronger relationship with the business than someone who purchased once.
A customer who has spent £10,000 deserves different commercial attention from someone who has spent £50.
RFM combines these three signals to produce a more complete picture of customer value. It is widely used to segment customers, understand where revenue comes from and create more relevant retention, loyalty and reactivation campaigns.
It is not a particularly new or sophisticated model. It does not require machine learning, a complicated attribution system or an expensive analytics platform. In its simplest form, all it needs is a customer identifier, a transaction date and a transaction value.
Yet it can reveal more about the health of your customer base than many impressive-looking dashboards.
RFM does not simply tell you who has spent the most.
That distinction is important.
Imagine two customers:
| Customer | Last purchase | Purchases | Total value |
| Customer A | 14 months ago | 1 | £10,000 |
| Customer B | 12 days ago | 8 | £8,000 |
If you sort customers only by revenue, Customer A appears more valuable.
RFM tells a different story.
Customer B has bought recently, has returned several times and has generated substantial revenue. Customer A may have completed a large one-off purchase and disappeared.
That does not make Customer A unimportant. It means the two customers need different treatment.
Customer B may belong in a loyalty, referral or expansion campaign. Customer A may require a reactivation campaign, a personal call or an investigation into why the relationship did not continue.
RFM turns a list of transactions into a map of customer relationships.
Most businesses know how many customers they have.
They know how much revenue they generated last month. They can probably tell you their average order value, conversion rate and cost per acquisition.
But they often struggle to answer more valuable questions:
Who are our best customers today?
Which customers are becoming more valuable?
Which loyal customers are quietly disappearing?
Who should receive an offer, and who does not need one?
Which customers look valuable because of one large transaction but are unlikely to return?
Many marketing campaigns treat everyone in the database as if they had the same relationship with the business.
A customer who purchased yesterday receives the same discount as someone who has been inactive for two years. A loyal customer receives the same onboarding message as a first-time buyer. A high-value client receives the same generic newsletter as everyone else.
This is not personalisation. It is distribution.
RFM creates a practical layer between customer data and marketing action. It helps a business decide:
The last point is frequently ignored.
Customer segmentation is not only about deciding whom to target. It is also about deciding whom not to target.
If someone purchases from you every month without an incentive, sending them a 30% discount may reduce your margin without changing their behaviour. The offer should perhaps go to a promising customer who needs encouragement to make a second purchase.
RFM helps separate customers who require an incentive from those who already demonstrate strong purchasing intent.
RFM is often presented as a method for sending better campaigns. That is useful, but it is not the main reason to perform the analysis.
There are two much more important reasons.
Your best customers should never become invisible simply because they are already buying from you.
Businesses often concentrate their attention on acquisition while assuming their established customers will continue purchasing. But customer relationships rarely disappear overnight. They usually deteriorate gradually.
A loyal ecommerce customer starts ordering less frequently. An agency client sends fewer briefs. A student stops completing lessons. A corporate training client does not book the next cohort.
Individually, these changes can look insignificant. RFM makes the pattern visible.
It helps you identify:
This is not about deciding which customers deserve good service. Everyone does.
It is about deciding how much attention, support and commercial investment each relationship requires.
A high-value client showing signs of disengagement might deserve a call from the founder. A promising new customer may need a better onboarding journey. A low-value customer repeatedly demanding disproportionate support may need clearer boundaries, a different service model or, occasionally, a polite exit.
Without this understanding, companies often distribute their attention according to who makes the most noise rather than who creates the most value.
RFM is not only useful for managing existing customers. It can also improve acquisition.
Once you identify your best customers, you can examine what they have in common:
This allows you to move beyond a vague ideal customer profile.
Instead of defining your target audience according to assumptions, job titles or demographics, you can define it using the observable characteristics of customers who have already proven their value.
You can then use those characteristics to:
This creates a valuable loop:
Identify your best customers. Understand what makes them different. Find more people like them.
RFM therefore helps you answer two sides of the same strategic question:
Whom should we protect, and whom should we pursue?
The first improves retention. The second improves acquisition.
Together, they help the company grow without assuming that every name in the database has the same value or the same potential.
The most common approach is to assign every customer a score from 1 to 5 for each dimension.
A score of 5 represents the most desirable behaviour, while 1 represents the least desirable.
For example:
| Days since last purchase | Score |
| 0–30 days | 5 |
| 31–60 days | 4 |
| 61–120 days | 3 |
| 121–240 days | 2 |
| More than 240 days | 1 |
| Number of purchases | Score |
| 10 or more | 5 |
| 7–9 | 4 |
| 4–6 | 3 |
| 2–3 | 2 |
| 1 | 1 |
| Total customer value | Score |
| Top 20% | 5 |
| Next 20% | 4 |
| Middle 20% | 3 |
| Next 20% | 2 |
| Bottom 20% | 1 |
A very strong customer might therefore receive an RFM score of 555.
Someone who purchased recently but has only bought once and spent relatively little might receive 511.
A previously valuable customer who has not returned for a long time might receive 155.
The individual digits are usually more useful than adding them together. A 155 customer and a 515 customer could have the same total score, but they describe completely different relationships.
The 155 customer was once frequent and valuable but is now inactive. The 515 customer purchased recently and spent a lot, but has bought only once.
They should not receive the same campaign.
There are two common ways to define RFM scores.
The first is to use fixed business rules, such as “a purchase in the last 30 days receives a Recency score of 5.”
The second is to divide the customer base into equal groups, usually quintiles. The top 20% receive a 5, the next 20% receive a 4, and so on.
Relative scoring is easy to automate and adapts to the data. However, it can sometimes create misleading distinctions. If almost everyone buys only once, dividing Frequency into five equal groups may not create commercially meaningful segments.
Fixed scoring is easier to connect to the natural buying cycle, but the boundaries need to be chosen carefully.
A supermarket, a car dealership, a university and a marketing agency cannot use the same definition of “recent.” Thirty days without a transaction could be alarming for a grocery delivery service and completely normal for a company selling executive education.
The scoring logic must reflect the expected customer journey.
Although RFM can theoretically create 125 combinations when each dimension is scored from 1 to 5, managing 125 campaigns would be impractical.
Most companies combine the scores into a smaller number of recognisable segments.
| Segment | Typical behaviour | Possible action |
| Champions | Recent, frequent and high value | Reward, involve and request referrals |
| Loyal customers | Frequent and consistently valuable | Retain, upsell and introduce loyalty benefits |
| Potential loyalists | Recent with early repeat behaviour | Encourage the next purchase |
| Promising | Recent but low frequency | Educate, nurture and recommend relevant products |
| New customers | Very recent first purchase | Improve onboarding and drive the second transaction |
| High-value newcomers | One recent, large transaction | Provide personal attention and identify expansion potential |
| At risk | Previously frequent or valuable, but inactive | Investigate and reactivate |
| Cannot lose them | Historically among the best, now disappearing | Use personal outreach rather than a generic email |
| Hibernating | Low Recency and limited historical activity | Use low-cost reactivation |
| Lost | Inactive with little historical value | Suppress or target selectively |
These labels are useful, but they are not universal truths.
The objective is not to reproduce someone else’s segment names. It is to create groups that lead to different decisions.
If two segments always receive the same treatment, you may not need two segments.
RFM should not remain inside a dashboard. You can use these segments directly in your advertising campaigns.
Start with your Champions, the customers with the highest Recency, Frequency and Monetary scores.
You can upload this segment to platforms such as Meta, Google or LinkedIn and use it as the source for a lookalike audience.
A lookalike audience is a new group of people who have not necessarily interacted with your business but share characteristics or behavioural patterns with your best existing customers. The advertising platform analyses the source audience and looks for other people who appear similar.
In simple terms, you are telling the platform:
“These are our best customers. Find us more people like them.”
The quality of the source audience matters. If you create a lookalike audience from everyone who has ever entered your database, the platform will look for more people who resemble an inconsistent mixture of excellent, average and poor customers.
If you use your Champions as the source, you are giving the platform a much clearer signal about the type of customer you want to acquire.
RFM can also be used to develop customers who already know your business.
Select your Potential Loyalists, the common RFM name for recent customers who have started showing repeat behaviour but have not yet become highly frequent or valuable. You can place this segment into a custom advertising audience and use it for retargeting.
Retargeting means advertising specifically to people who already know your business. They may be previous buyers, leads, website visitors, subscribers or people who have interacted with your content. Instead of trying to introduce the business to a completely new audience, you are encouraging a known person to take the next step.
For example, you could show Potential Loyalists:
The objective is not simply to generate another sale. It is to help a promising customer develop into a loyal one.
The distinction is useful:
You can apply the same logic to other RFM segments.
Customers who are At Risk can receive a reactivation campaign. High-value newcomers can receive messages designed to create a second purchase. Loyal Customers can see referral, membership or premium-product campaigns. Lost Customers can be excluded from expensive campaigns or targeted only through lower-cost channels.
This connects RFM directly to both sides of growth: acquiring more customers who resemble your best ones and developing existing customers who have already shown signs of future value.
At a minimum, you need a transaction table containing:
Useful additional fields include product category, gross margin, channel, location, salesperson, acquisition source and refund value.
For each customer, you calculate:
[
\text{Recency} = \text{Analysis date} – \text{Date of last valid transaction}
]
[
\text{Frequency} = \text{Number of valid transactions during the analysis period}
]
[
\text{Monetary value} = \text{Sum of valid transaction values during the period}
]
The basic output might look like this:
| Customer ID | Last transaction | Recency days | Frequency | Monetary value |
| C001 | 10 Aug 2026 | 12 | 8 | £8,450 |
| C002 | 3 Mar 2026 | 172 | 3 | £2,100 |
| C003 | 20 Aug 2026 | 2 | 1 | £6,800 |
This can be extracted from an ecommerce platform, CRM, invoicing system, learning management system or data warehouse.
The process is usually:
RFM is not a report you create once.
Customers move between segments. A Champion can become At Risk. A first-time customer can become a Potential Loyalist. A dormant customer can return.
Monthly recalculation may be enough for an agency or educational institution. An active ecommerce shop may update the scores daily.
The mathematics is the easy part. Data definitions create most of the problems.
Should two orders placed on the same day count as two purchases or one purchasing occasion?
Should Frequency represent orders, subscriptions, invoices or completed projects?
Should Monetary value use gross revenue, net revenue or gross profit?
Should refunded orders be removed entirely?
Should purchases from two email addresses be combined if they belong to the same person?
Should a company with ten contacts be analysed as one account or ten individuals?
There is no universally correct answer. There must, however, be a consistent one.
An agency may learn more from gross profit than revenue. An educational institution may want to count completed courses rather than registrations. An ecommerce company may need to subtract returns before calculating customer value.
RFM is only as useful as the business definitions underneath it.
RFM works particularly well when transactions are frequent enough to reveal a pattern.
But not every valuable relationship produces frequent purchases.
Someone may read every article, attend three webinars, participate in a community and use a free tool without having bought anything yet. From an RFM perspective, this person has little or no value. From a commercial perspective, they may be one of the strongest future opportunities.
This is where RFE becomes useful.
RFE stands for:
RFE is a variation of RFM designed for environments where attention, participation or product usage can be more informative than immediate spending.
The exact definition of Engagement depends on the business. It might include visits, downloads, product usage, event attendance, lesson completion, community participation or interactions with the sales team.
The important part is not to treat every action as equal.
Opening an email should not carry the same weight as attending a two-hour webinar. Visiting a pricing page should probably be worth more than liking a social post. Submitting an assignment is a stronger educational signal than opening a lesson.
A simple Engagement score might look like this:
| Action | Engagement points |
| Email open | 1 |
| Article visit | 2 |
| Resource download | 4 |
| Webinar registration | 5 |
| Community contribution | 8 |
| Webinar attendance | 10 |
| Product demo request | 15 |
| Referral | 20 |
RFE can identify engaged prospects before they purchase, active users who may be ready to upgrade and existing customers whose declining activity could signal future churn.
You do not necessarily need to choose.
RFM measures commercial behaviour. RFE measures behavioural engagement.
Together, they can reveal situations that neither model would identify alone.
| RFM | RFE | Interpretation |
| High | High | Valuable and actively connected |
| High | Low | Valuable, but engagement may be declining |
| Low | High | Strong future potential or a non-paying advocate |
| Low | Low | Limited current value and limited intent |
A more complete model can be called RFME: Recency, Frequency, Monetary value and Engagement.
But more dimensions do not automatically produce better decisions. Begin with RFM. Add Engagement only when you know what behaviours matter and what action a change in the score should trigger.
Ecommerce is the most natural environment for RFM because purchases are usually structured, timestamped and connected to customer accounts.
These are not simply the customers with the largest historical spend. They are people who purchased recently, return regularly and generate meaningful value.
You can give them early product access, invite them into a VIP programme, ask for reviews or referrals and avoid wasting discounts on behaviour they would have demonstrated anyway.
You can also use them as the source for lookalike audiences, allowing advertising platforms to find new prospects who share characteristics with the customers who already create the most value.
A customer with a score such as 515 may have placed one large recent order.
The objective is not to congratulate yourself on the revenue. It is to create the second purchase.
Analyse what they purchased, recommend the natural next product and provide service that reduces the risk of the relationship remaining a one-off transaction.
Potential Loyalists have purchased recently and are beginning to demonstrate repeat behaviour, but they have not yet developed into Champions or Loyal Customers.
This is an ideal segment for retargeting. You can show them complementary products, a carefully chosen incentive or a message designed around the next natural step in their journey.
A customer may still have strong Frequency and Monetary scores while their Recency score falls.
This is one of the most valuable RFM signals. Revenue dashboards often continue to show this person as a top customer because historical spending remains high. RFM reveals that the relationship is becoming inactive.
Compare customers acquired through heavy discounts with organically acquired customers.
If discounted customers have strong initial Recency but weak long-term Frequency, the promotion may be generating orders rather than customers.
Add the first purchased product or product category to the analysis.
You may discover that one product attracts high-volume first-time buyers but generates little repeat business, while another lower-volume product consistently creates loyal customers.
This can influence acquisition budgets, bundles, merchandising and content strategy.
RFE can identify customers who repeatedly browse products, use wish lists, open back-in-stock notifications or visit high-intent pages without purchasing.
That group may need reassurance, availability, better payment options or clearer shipping information rather than another generic discount.
RFM requires adaptation for agencies because purchases are less frequent, contracts last longer and a large invoice does not necessarily mean a profitable relationship.
For an agency, the dimensions might become:
Gross profit is often a better signal than revenue. A £100,000 client requiring excessive delivery resources may be less valuable than a £60,000 client with strong margins, easy collaboration and referral potential.
High-revenue clients are not automatically strategic clients.
A strong agency client is current, repeatedly assigns work, generates healthy commercial value and engages positively with the team.
RFM can reveal which accounts deserve senior attention, proactive ideas and account expansion plans.
Agencies often discover churn too late.
There may be no formal cancellation. The briefs simply become less frequent. The client stops attending planning sessions. Response times increase. New work is postponed.
Declining Recency and Frequency can provide an early warning before the revenue disappears completely.
A recent, high-value, low-frequency client is a clear expansion opportunity.
Instead of adding them to a generic nurture campaign, the agency can review the original project, identify the next logical need and prepare a specific recommendation.
Add the first service purchased to the analysis.
Do clients entering through SEO stay longer than those entering through a website redesign? Do strategy engagements lead to more cross-selling? Do training clients later purchase consulting?
This helps the agency understand which entry services create the strongest lifetime relationships.
Once you identify your Champion clients, study their industries, company sizes, commercial challenges, entry services, decision-makers and acquisition sources.
You can then use these characteristics to refine your ideal customer profile, prospecting lists and advertising audiences. On platforms that support it, you may also use these clients as the source for a lookalike audience.
For agencies with a relatively small number of clients, the RFM data may need to be combined with website visitors, qualified leads or other high-intent signals to create a large enough advertising audience.
RFE can incorporate operational engagement such as:
A client may have a high RFM score but declining Engagement. That can signal account risk, even while invoices are still being paid.
Conversely, a smaller client with high Engagement may have significant growth, referral or case-study potential.
Educational institutions often focus on enrolment. But enrolment is only the beginning of the relationship.
RFM can be used to understand learners, corporate clients, alumni, donors or programme sponsors.
For individual learners:
For corporate education clients:
A recent learner who has completed several related courses is a strong candidate for an advanced programme, certification or specialisation.
The recommendation should follow the learning journey, not simply promote whichever course needs more enrolments.
A learner who recently completed one course should receive a thoughtful progression path.
What should they learn next? What level are they now ready for? Which skill would complement what they have already completed?
RFM helps turn a course catalogue into a customer journey.
A previously frequent learner with a low Recency score may be ready for reactivation through an alumni event, a new advanced programme or updated content in their field.
The message can acknowledge their history instead of treating them as a new lead.
Analyse the first programme taken by students who later become frequent, high-value learners.
Some introductory courses may generate many enrolments but little continuation. Others may consistently lead to advanced programmes, memberships or corporate training.
This helps the institution evaluate programmes by the relationships they create, not only by immediate enrolment revenue.
Create a Champion segment from learners who enrol recently, return for additional programmes and generate meaningful value.
You can use this audience as the source for lookalike campaigns designed to find new people who resemble your most committed students.
Potential Loyalists can then become a retargeting audience. These may be learners who completed an introductory course, attended a recent webinar or made an initial purchase but have not yet progressed to the next stage.
The message should not simply say, “Buy another course.” It should show them the next logical step in their learning journey.
This is where RFE becomes essential.
Educational Engagement can include:
A student can have a high Monetary score and a very low Engagement score. Commercially, the enrolment looks successful. Educationally, the student may be at risk.
An institution that identifies this early can intervene with reminders, tutoring, cohort support or a different learning path.
A learner may attend free webinars, contribute to the community, complete free lessons and interact with faculty without having purchased a major programme.
RFM would underestimate this person. RFE may identify them as a strong future student, ambassador or corporate introduction.
At the same time, engagement needs to be interpreted carefully. Some people genuinely value free educational resources but may never intend to buy. RFE helps identify interest, but it does not automatically prove commercial intent.
An RFM score is useful. A change in RFM score is usually more useful.
A customer moving from 555 to 455 may still appear healthy, but Recency has started to decline.
A customer moving from 311 to 423 is developing into a repeat buyer.
A learner moving from high Engagement to low Engagement may need support.
An agency client whose meeting attendance and project frequency are both falling may be preparing to leave.
Instead of looking only at the current segment, track:
This turns RFM from a static customer report into an early-warning and opportunity-detection system.
RFM uses past behaviour to create a useful interpretation of the present. It does not explain why a customer behaved that way, and it cannot guarantee what they will do next.
A customer may appear inactive because their purchase cycle is naturally long. A high-spending customer may have terrible margins. A highly engaged learner may have no intention of purchasing. A frequent agency client may be generating operational complexity that makes the account unprofitable.
RFM should therefore be combined with context, customer research and business judgement.
It is also not a replacement for customer lifetime value, churn prediction or cohort analysis. It is often the best place to start because it is simple, explainable and immediately actionable.
More advanced models can be added once the organisation has learned how to act on the basic signals.
The temptation is to build a beautiful RFM dashboard with colourful segments and impressive charts.
That is not the objective.
Before building the analysis, ask:
What will we do differently when someone enters this segment?
If there is no answer, the segment has little operational value.
A useful RFM system should connect each important change to an action. That might be an automated email, a CRM task, an account review, a personal call, a recommendation, an educational intervention, a retargeting campaign or an exclusion from a discount audience.
RFM is valuable because it converts transactions into relationships, and relationships into decisions.
It tells you whom to protect, whom to develop, whom to reactivate, whom to pursue and whom to stop pursuing.
The analysis itself is relatively simple.
What matters is what you notice early enough to do something about it.
Theodore has 20 years of experience running successful and profitable software products. In his free time, he coaches and consults startups. His career includes managerial posts for companies in the UK and abroad, and he has significant skills in intrapreneurship and entrepreneurship.
Built an AI routine that checks who is emailing me, understands the business context, filters…
Traditional online communities were built for a different internet. Community 2.0 brings humans and AI…
MCP-first apps change how software is built and used. Discover why the UI is becoming…
A step-by-step methodology for measuring how often AI systems cite your brand, with the query…
In 2026, growth hacking is no longer about finding single-turn shortcuts; it has evolved into…
Accessing AI data is only half the challenge. Learn how to transition from static Markdown…