RFM Analysis: How to Find Your Best Customers (and What to Send Them)

Somewhere in your customer list is a relatively small group of people doing an outsized share of the work.

They buy more often. They spend more. They come back sooner. And when they start to drift, the revenue risk adds up fast.

RFM analysis is one of the fastest ways to find them.

It’s a segmentation method that has been around since the catalog era, and it still works because it’s built on the three purchase behaviors that predict future buying better than almost anything else: how recently someone bought, how often they buy, and how much they spend.

Done well, RFM helps you find your best customers, spot strong repeat buyers before they become obvious, and identify the people quietly slipping away before their next order never happens.

In one apparel store we analyzed, a strict “can’t lose them” segment contained 15,108 lapsed former high-value customers, representing $4.8M in historical revenue already at risk.

This guide covers how RFM works, the segments worth building, what to send each one in Klaviyo, and where RFM starts to fall short.

What is RFM analysis?

RFM stands for Recency, Frequency, Monetary value.

Each customer gets scored on all three, usually on a 1–5 scale:

• Recency (R): How recently the customer purchased. A customer who bought last week scores 5; one who has not ordered in a long time scores 1.
• Frequency (F): How often they purchase. Someone on their sixth order scores much higher than a one-time buyer.
• Monetary (M): How much they have spent in total. Your top spenders score 5.

Combine those three scores and every customer lands in a segment with an obvious next move.

A 5-5-5 is a champion you should not be training to wait for discounts. A 1-5-5 is the opposite problem: a former best customer who has gone quiet.

In the store above, some champion customers had already placed 13–14 orders and spent $1.4k–$1.6k, while still scoring at the top end on recency. By contrast, top “can’t lose them” customers had also spent $1.3k–$1.5k across 11–13 orders, but had gone 540–742 days without buying again.

That is the core logic behind the RFM model: past buying behavior is still the strongest simple signal of what a customer is likely to do next. Someone who bought multiple times recently is far more likely to buy again than someone who purchased once a long time ago, even if their order values were similar.

How to calculate RFM scores

You need three numbers per customer, all sitting in your order data:

• Date of last order → rank customers into quintiles, most recent = 5
• Total number of orders → quintiles again, most frequent = 5
• Total spend → quintiles again, highest spend = 5

The key is to divide your customer base into five equal groups for each metric instead of using fixed thresholds like “purchased in the last 30 days.”

Why? Because quintiles adapt to your store’s actual buying rhythm.

A 45-day gap means something very different for a replenishment brand than for a durable-goods brand. A “high-frequency” customer in one store might look average in another. Quintiles let the store define “good” and “bad” relative to its own customer base.

You can do the math in a spreadsheet. The harder part is keeping it current.

RFM scores get stale quickly. A customer can move from “loyal” to “at risk” in the weeks between manual refreshes, which is often exactly when a well-timed email could have saved the next order.

The RFM segments that matter

You do not need all 125 possible score combinations. In practice, a handful of segments cover most of the revenue decisions an ecommerce brand actually needs to make.

Segment	Typical scores	Who they are
Champions	R 4–5, F 4–5, M 4–5	Recent, frequent, high-spend customers
Loyal repeat buyers	R 3–5, F 4–5	Consistent repeat buyers, not always top spenders yet
Potential loyalists	R 4–5, F 2–3	Recent second- or third-order customers
New customers	R 5, F 1	First-time buyers in the highest-leverage post-purchase moment
At risk	R 1–2, F 3–5	Previously active customers starting to drift
Can’t lose them	R 1, F 4–5, M 4–5	Former top customers who have lapsed

Here is what that looked like in one store:

Segment	Customers	Avg. days since last order	Avg. orders	Avg. spend per customer	Historical revenue
Champions	79,223	204 days	2.8	$357	$28.3M
Loyal repeat buyers	44,654	370 days	2.5	$297	$13.3M
Potential loyalists	56,779	200 days	1.0	$137	$7.8M
New customers	14,165	154 days	1.0	$135	$1.9M
At risk	27,321	712 days	1.0	$95	$2.6M
Can’t lose them	61,342	645 days	2.2	$279	$17.1M

That is why RFM is useful: it turns a customer table into a set of groups with clearly different economics and clearly different messaging needs.

What to send each segment in Klaviyo

RFM only matters if each segment gets treated differently.

This is where Klaviyo is useful: once the segments exist, you can send the right message to the right customer at the right time.

Champions

Send:

• Early access to launches
• Referral requests
• UGC asks
• VIP-only perks

Keep them out of heavy discount campaigns. They are your best customers already. If they would have paid full price, the discount just gives away margin.

Loyal repeat buyers

Send:

• Loyalty rewards
• Exclusive bundles
• Cross-sells based on what they already own

These customers have already shown repeat behavior. The job is to deepen the relationship and widen the basket.

Potential loyalists

Send:

• A third-purchase incentive
• Bundle recommendations
• Subscription or replenishment offers where relevant

This is often the highest-leverage segment in the file. They are not proven long-term yet, but they are close.

New customers

Send:

• Product education
• “Get the most from your purchase” content
• Review requests
• Cross-sell messages timed to when second orders typically happen

Your first post-purchase window is usually the cheapest and most controllable second-order opportunity you will get.

At risk

Send:

• Personalized product recommendations
• New arrivals tied to prior purchases
• Reminders timed before the expected reorder window, not long after it

Timing matters here. The goal is to intercept drift before it turns into full lapse.

Can’t lose them

Send:

• A founder-style email
• A genuinely exclusive offer
• Concierge treatment
• Messaging that acknowledges their prior value

For this group, “you have been one of our best customers” often lands better than a generic sitewide discount.

In the store above, the strictest version of this segment alone held 15,108 customers with $4.8M of historical revenue behind them. That is why this segment deserves special handling.

Where RFM analysis falls short

RFM is a strong starting point. But it has real blind spots.

1. It is entirely backward-looking

Recency, frequency, and monetary value all describe what already happened.

RFM can tell you a customer was valuable. It cannot tell you who is about to lapse before the recency score catches up. By the time a customer fully “looks at risk” in a simple backward-looking model, some of the intervention window may already be gone.

2. It treats every dollar as equal

Two customers can both spend $500 and get the same monetary score.

But one may have come in through paid acquisition, bought discounted low-margin products, and never generated much contribution profit. The other may have arrived organically, bought at full price, and produced far more profit for the business.

Same revenue score. Very different value.

3. It ignores the context around the purchase

RFM does not know:

• what someone bought first
• which channel they came from
• whether they are a subscriber
• which product paths lead to stronger long-term value

That matters because customer value often depends on more than just how much and how often they have spent so far.

In this same store, 25.8% of customers whose first order was Accessories later bought Outerwear, while 23.0% of customers whose first order was Outerwear later bought Accessories. Customers whose first order was Sweatshirts later moved into Joggers at a 28.5% rate.

A plain RFM score will not surface those product-path differences, even though they can matter a lot for retention and merchandising.

Beyond RFM: adding profit and prediction

None of this means abandoning RFM.

It means enriching it with the data your email platform usually does not hold natively: full-order-history value, profitability, and each customer’s likely future behavior.

That is the layer the Profit Agent adds on top of Klaviyo. It’s built on Lifetimely’s purchase and profitability data across $100B+ in GMV from Shopify stores, so instead of scoring customers only on what they did, you can also segment on:

• predicted 12-month LTV
• predicted profit
• repurchase timing
• true lapse risk
• which first products and paths create stronger future customers

In the store above, 19,230 customers already have predicted-value coverage, with an average predicted 12-month LTV of $581 and average predicted 12-month profit of $568.

And at the individual customer level, some customers with only $561–$671 in realized revenue already show $743–$750 in predicted 12-month LTV. That is exactly the kind of signal RFM misses: customers who may not yet rank in the top revenue quintile, but are still worth treating like future VIPs.

That is how you find:

• Hidden champions: customers with stronger future value than current spend suggests
• True at-risk customers: customers drifting past their own normal reorder rhythm, not just a generic recency cutoff
• Higher-value acquisition targets: the first products and customer paths that produce better long-term buyers

How the two work together

The simplest way to think about it:

• RFM gives you a fast behavioral segmentation framework
• Klaviyo turns those segments into campaigns and flows
• Profit Agent improves the segments by adding profit, prediction, and customer-value context

So instead of asking only, “Who bought recently, often, and a lot?”, you can also ask:

• Which of my high-value customers are slipping before they fully lapse?
• Which recent buyers look likely to become top customers?
• Which first-purchase paths create the most valuable repeat customers?
• Which customers are worth protecting because they drive profit, not just revenue?

And instead of exporting data, scoring it manually, and re-importing lists, you ask in plain language: “Which of my VIPs haven’t purchased in a while and are starting to slip away?” The Profit Agent finds them in your purchase and LTV data, shows you what the group is worth, and builds it as a live list or segment directly in Klaviyo, named, sized, and ready to send to. The connection stays in sync both ways, so your segments update as customer behavior changes instead of going stale between manual refreshes.

Start with your own data

RFM analysis will take you a long way.

Score your customers. Build the six core segments. Treat each one differently. That alone will put you ahead of most brands still sending the same campaign to everyone.

Then, when you are ready to go beyond backward-looking purchase history, add the layers that RFM cannot see on its own: profit, predicted value, repurchase rhythm, and first-product context. Connect Klaviyo to the Profit Agent and build your first value-based segment from a single question.

Your best customers are already in your data.