RFM Matrix: Segment Customers with Real Data

80% of a company's future revenue will come from 20% of its current customers (Gartner, 2025). The question is: do you know who that 20% is?

The RFM matrix is one of the most effective ways to answer that question. It's been used in direct marketing for decades and, with the data any ecommerce generates today, implementing it is more accessible than ever. This article explains what it is, how scoring works and how to apply it to improve conversion.

What is RFM analysis

RFM is a customer segmentation model based on three purchase behavior variables:

Recency: when the customer last made a purchase
Frequency: how many times they've purchased in a given period
Monetary: how much money they've spent in total

The premise is simple: a customer who bought recently, buys frequently and spends a lot is more valuable than one who bought a year ago, only once and for a small amount. It sounds obvious, but most companies treat both customers the same way.

Why it works

The RFM model works because it's based on actual behavior, not predictions or demographic data. You don't need to know the customer's age, gender or interests. You only need three data points you already have in your transactional database:

Date of last purchase
Number of purchases
Total amount spent

How RFM scoring works

Step 1: Collect the data

For each customer in your database, you need:

Variable	Required data	Source
Recency	Date of last purchase	Order database
Frequency	Total number of completed orders	Order database
Monetary	Sum of all order amounts	Order database

Step 2: Divide into quintiles

For each variable, rank all customers and divide them into 5 equal groups (quintiles). Assign a score from 1 to 5:

Recency:

5 = purchased in the last 7 days
4 = purchased in the last 30 days
3 = purchased in the last 90 days
2 = purchased in the last 180 days
1 = purchased more than 180 days ago

Frequency:

5 = 10+ purchases
4 = 6-9 purchases
3 = 3-5 purchases
2 = 2 purchases
1 = 1 purchase

Monetary:

5 = top 20% in spending
4 = 20-40%
3 = 40-60%
2 = 60-80%
1 = bottom 20%

Note: ranges are indicative. Each business should define them based on their own reality (average order value, typical purchase frequency, seasonality).

Step 3: Combine the scores

Each customer receives a 3-digit code. For example:

555: purchased recently, buys frequently, spends a lot. Your best customer.
111: purchased long ago, only once, and for little. Customer at risk of total loss.
551: purchased recently, buys frequently, but spends little. Upselling opportunity.
155: hasn't purchased in a long time, but when they did it was frequent and high-value. Customer to recover urgently.

Step 4: Create actionable segments

You don't need to work with 125 combinations (5x5x5). Group the scores into business-meaningful segments:

Segment	RFM score	Characteristics	Recommended action
Champions	555, 554, 545	Buy a lot, frequently and recently	VIP program, exclusivity, referrals
Loyal	435, 534, 443	Buy frequently with good value	Upselling, cross-selling, early access
Potential loyals	512, 513, 412	Purchased recently with medium frequency	Incentivize second/third purchase
New	511, 512, 411	Recent first purchase	Onboarding, welcome email
At risk	255, 254, 245	Were good customers but haven't purchased in a while	Reactivation campaigns, special offers
Hibernating	155, 144, 133	High historical value but inactive	Aggressive win-back or accept the loss
Lost	111, 112, 121	Low value and no activity	Exclude from paid campaigns, database cleanup

Ecommerce CRO applications

RFM analysis isn't just for email marketing. It has direct applications in optimizing your website's conversion.

1. Experience personalization

If you know the RFM segment of a returning user (via login or cookie), you can personalize their experience:

Champions: show premium products, early access to launches
New: highlight bestsellers, usage guides, guarantees
At risk: pop-up with recovery offer, products they previously bought at a discount
Potential loyals: recommendations based on their history to encourage the second purchase

2. Retargeting campaign optimization

Instead of retargeting all visitors equally, segment by RFM:

Champions: they don't need a discount. Show them what's new.
At risk: an aggressive discount is justified. Acquiring a new customer costs 5-7x more than retaining an existing one.
Lost: exclude this segment. Retargeting won't bring them back, and you're wasting budget.

3. Marketing budget allocation

Where should you invest more? RFM gives you the data-driven answer:

Segment	Recommended investment	Why
Champions	Medium (retention is cheap)	They're already loyal. Maintain the relationship
Potential loyals	High	Maximum ROI from converting to loyal
At risk	Medium-high	Recovering a customer costs less than acquiring a new one
New	High	The first post-purchase experience defines the relationship
Lost	Low or none	Low expected ROI

4. A/B test design

RFM can inform your CRO experiments. There's no point testing the same variant for Champions and New customers: their motivations and behaviors are radically different.

Examples of RFM-segmented tests:

Champions: referral program test vs. no program
New: long vs. short onboarding email test
At risk: 10% discount vs. 20% discount vs. free shipping test
Potential loyals: checkout cross-selling vs. post-purchase cross-selling test

Tools for implementing RFM

Native solutions in ecommerce platforms

Klaviyo: integrated RFM segmentation with automatic scoring. The most accessible option for Shopify ecommerce.
Omnisend: customer segmentation with RFM-type behavioral variables.
Drip: contact scoring based on purchase behavior.

Analytics tools

Google BigQuery + Looker Studio: for ecommerce with data in BigQuery, you can build RFM scoring with SQL and visualize it in dashboards.
Python (pandas + scikit-learn): the most flexible option. RFM scripts with K-means clustering for advanced segmentation.
Excel / Google Sheets: perfectly viable for databases of up to 50,000 customers. Quintiles using the PERCENTILE function.

CDPs (Customer Data Platforms)

Segment: customer data collection and unification for real-time RFM scoring.
Bloomreach: experience personalization based on native RFM segmentation.

Practical example: fashion ecommerce

Imagine a fashion ecommerce with 20,000 customers. After applying RFM analysis:

Segment	% of customers	% of revenue	Action
Champions (555-445)	8%	42%	VIP program, permanent free shipping
Loyal (434-335)	15%	28%	Cross-selling via email
Potential loyals (512-312)	12%	10%	Second purchase discount
New (511-411)	20%	8%	Optimized welcome sequence
At risk (244-145)	18%	9%	Win-back campaign with discount
Lost (111-122)	27%	3%	Exclude from paid campaigns

The most revealing data point: 8% of customers generate 42% of revenue. Without RFM, you'd be investing the same to retain your Champions as to try to recover lost customers who represent just 3% of your revenue.

Limitations of the RFM model

Like any model, RFM has limitations you should be aware of:

It describes, not predicts. RFM tells you how a customer behaves today, not how they'll behave tomorrow. For prediction, you need models like predictive CLV or churn probability.
It doesn't account for margin. A customer who spends a lot but only buys discounted products may have a real value lower than their Monetary score suggests.
Seasonality. A customer who only buys on Black Friday might appear as "At risk" in July. Adjust Recency ranges according to your business's seasonality.
One-time purchase products. If you sell mattresses or large appliances, Frequency has a different meaning. Adapt the model to your natural purchase cycle.

The RFM matrix is a powerful tool because it turns data you already have into actionable decisions. You don't need machine learning or sophisticated tools to get started: a spreadsheet and your order database are enough.

If you want to go beyond segmentation and optimize the conversion experience for each segment, at Boost we do data-driven CRO. And for a first diagnostic of your website, try Scan&Boost.

— Adrià Vidal