Nearly every Klaviyo account has a segment called VIPs. Usually it's "spent over £300" or "placed three or more orders", built once, never revisited.
Somebody will tell you that's RFM segmentation. It isn't. It's one condition with a nice name on it.
Real RFM splits your customer base into groups that behave differently and need different messages. It's one of the few pieces of genuinely advanced segmentation that doesn't require a data team, and it's the thing that stops you sending the same email to somebody who bought last week and somebody who bought in 2023.
The three dimensions, and why one condition can't replace them
RFM stands for recency, frequency and monetary value. How long ago did they buy, how often do they buy, how much do they spend.
The reason all three matter is that any one of them in isolation gives you a misleading picture.
Take two customers who've each spent £500 with you. The first spent it in one order two years ago. The second spent it across twelve orders in the last six months.
Identical monetary value. Completely different people. One is a lapsed one off, the other is among the most valuable customers you have. A segment built on total spend puts them in the same bucket and sends them the same email.
Add recency and frequency and the two of them separate on the spot. That separation is what you're paying for.
What Klaviyo gives you, and what it costs
Klaviyo does have a native RFM report. Two things to know before you go looking for it.
It sits inside Advanced KDP or Marketing Analytics, which are paid additions on top of the standard platform rather than included features. And it requires at least 500 customers who have placed an order, which again means purchasers rather than profiles.
The native report scores each dimension on a 1 to 3 scale and combines them into groups with names like Champions and At Risk. It's well built and if you've got it, use it. Klaviyo exposes the resulting group as a profile property, so you can segment and trigger flows off it directly.
The limitation is granularity. A 1 to 3 scale across three dimensions gives you 27 possible combinations, and in practice they collapse into a handful of named groups. For a brand with a broad range of customer values, that's coarse. Most external RFM frameworks score 1 to 5, which gives 125 combinations and far more room to separate a good customer from an exceptional one.
If you're not paying for the add on, or you want finer control, build it yourself. Everything you need is in the base segment builder.
Building it by hand
You're constructing three scores and then combining them.
Recency. Work out the distribution of days since last order across your customer base, then split it into bands. Don't use round numbers, use your actual data. For a brand with a 30 day purchase cycle, 90 days is lapsed. For one with a 200 day cycle, 90 days is completely normal. This is why calculating your median purchase latency first makes everything else easier.
Frequency. Number of orders placed. Usually simpler, because the distribution is steep. One order, two to three, four plus is a workable starting split for most brands.
Monetary. Total spend, banded into quintiles across your purchaser base. Use quintiles rather than round numbers so the bands reflect your actual customer distribution rather than what sounds tidy.
Then build segments combining them. You don't need all 125 combinations. You need the eight or so that you'd actually treat differently.
The groups that earn their keep
Champions. Recent, frequent, high spend. They don't need discounts. They need early access, new product first, and to feel like insiders. Discounting this group is pure margin destruction, because they were going to buy regardless.
Loyal but lower spend. Frequent and recent, smaller basket. The opportunity is basket size, not frequency. Bundles, upgrades, free shipping thresholds set just above their typical order.
High spend, low frequency. Big infrequent orders. Don't panic when they go quiet, that's their normal pattern. Time your contact to their cycle rather than your calendar.
Recent one off buyers. Bought once, bought recently. The most important group in the whole framework, because this is where second purchase odds are decided and the window is short. They get education and a well timed second purchase push, not a generic newsletter.
Slipping away. Previously frequent, now overdue relative to their own pattern. This is the group where intervention actually works, and it's the group most brands miss because they define lapsed by a fixed number of days rather than relative to that customer's history.
At risk high value. Slipping away and they used to spend a lot. Deserve individual attention. For a small enough group, a personal email from a real person beats any automation.
Lapsed. Long gone, low frequency. Worth one honest winback attempt, then suppression. Don't spend your domain reputation on people who've stopped answering.
New. No purchase history to score. They belong in the welcome and post purchase programme, not in RFM analysis.
What it looks like with real numbers
Abstract frameworks are easy to nod at and hard to act on, so here's a worked version.
Brand doing £2m a year, 12,000 purchasing customers, average order £65.
Run the bands and you might find something like: 600 Champions averaging £680 lifetime spend. 1,400 Loyal but lower spend, averaging £210. 5,800 Recent one off buyers at £65. 1,900 Slipping away, previously good, now overdue. 300 At risk high value, averaging over £500 and drifting. The rest lapsed.
Two things jump out of that table immediately.
The 600 Champions represent about £408,000 of historic revenue from 5% of the customer base. Every blanket 20% off campaign you send costs you real money against that group, because they'd have bought anyway.
The 300 At risk high value customers represent around £150,000 of historic spend and they're leaving. Three hundred people is a small enough number that a human being could email each of them personally over a fortnight. It never happens, because it isn't automated and it doesn't scale. Which is precisely why it would work.
That second observation is usually worth more than the entire rest of the exercise, and you can't see it without the segmentation.
The mistake that costs the most money
It's discounting the wrong groups.
The default behaviour in most accounts is that discounts go to whoever is on the list, which means your Champions receive every promotion you send. They'd have bought anyway. You've just paid them not to.
Meanwhile the group that genuinely needs an incentive, engaged people who haven't converted, or high value customers starting to drift, gets the same 15% off as everyone else because there's no mechanism to treat them differently.
Discount depth should be inversely related to purchase probability. The more likely somebody is to buy without an incentive, the less incentive they should receive. That's a sentence most brands would agree with in principle and almost none have implemented, because implementing it requires exactly this kind of segmentation.
Proving it worked
Build this and then measure it properly, or you'll end up defending it on open rates like everybody else.
Three numbers, tracked before and after.
Revenue per recipient by group. The whole point of segmentation is that a smaller, better matched send earns more per person. If revenue per recipient hasn't moved, the segmentation isn't doing anything.
Discount spend as a share of revenue. This should fall. You're withdrawing discounts from people who didn't need them. If it hasn't fallen, you've built the segments and carried on sending everyone the same offer, which is where most implementations quietly end up.
Movement between groups. How many Recent one off buyers became Loyal this quarter. That's the retention machine working, and it's the number that connects segmentation to lifetime value rather than to this month's send.
If you want to be certain the groups are doing the work rather than the calendar, hold out a portion of one group and compare. Same principle as testing a flow. Uncomfortable, cheap, and it settles the argument permanently.
Keeping it alive
RFM built once and left alone rots. Customers move between groups constantly, and a static segment built in March describes a customer base that no longer exists by September.
Build it with dynamic segment conditions so membership updates automatically, then review the band definitions quarterly. As your business changes, the thresholds that separate a high spender from a mid spender change too.
And measure the right thing. RFM is judged on repeat purchase rate, revenue per recipient, and the margin you keep by not discounting people who did not need it. Open rates tell you nothing about any of that.
Start with three groups if eight feels like too much: Champions, Recent One Off Buyers, and Slipping Away. Those three cover most of the money. Add the rest once you've proved the first three do something.
And if anyone tells you they have already done RFM on your account, ask to see the group definitions and how often membership refreshes. If the answer is a single segment called VIPs, built two years ago and never touched, you now know what to ask for next.