- Classic RFM (recency, frequency, monetary value) tells you who your valuable customers are, but not who is ready to buy this week. On its own it is a rear-view mirror.
- Our Value × Intent (RFMB) model adds a Buying Stage layer from the last 30 days: B1 visit, B2 product view, B3 add to cart, B4 checkout, B5 purchase. Value decides how much you can afford to spend on a customer; intent decides when.
- Build three value tiers and five intent stages in your ESP, then write rules for the handful of combinations that change what you send, especially who should never get a discount.
What is RFM segmentation, and why isn’t it enough on its own?
RFM analysis scores each customer on three questions:
- Recency: How long since their last order?
- Frequency: How many orders have they placed?
- Monetary value: How much have they spent, net of refunds?
Recent, frequent, high-spending customers are your best. Frequent buyers who have gone quiet are at risk. That much is standard audience segmentation, and it beats sending everyone the same campaign.
The limit is that RFM only looks at orders. It cannot see that a lapsed VIP browsed your new collection this morning, or that a one-time buyer has had the same product in their cart for three days. Two customers with identical scores can be in completely different moments.
How do you score RFM for a Shopify store?
A spreadsheet and a customer export are enough. Aim for a score from 1 to 3 on each dimension, which is enough resolution for most stores.
Step 1: Pull clean order data
Export customers with order count, total spent and last order date from Shopify. Before scoring:
- Use net revenue after refunds, not gross sales.
- Remove test orders, staff accounts and wholesale or B2B buyers, who distort frequency and value.
- Merge obvious duplicates, such as the same person checking out as a guest under two emails.
Step 2: Set thresholds that match how your customers buy
Don’t split every dimension into equal fifths by default. Recency should follow your repurchase cycle; a coffee brand and a furniture brand cannot share windows. Frequency works best in fixed bands, because the large group of one-time buyers breaks percentile splits.
Here is an illustrative scoring grid for a store whose typical customer reorders within about 60 days:
| Score | Recency (days since last order) | Frequency (orders) | Monetary (net spend) |
|---|---|---|---|
| 3 | 0 to 60 | 4 or more | Top third of customers |
| 2 | 61 to 180 | 2 to 3 | Middle third |
| 1 | More than 180 | 1 | Bottom third |
If your margins vary a lot by product, score monetary value on contribution margin instead of revenue. A customer who spends $400 on low-margin bundles may be worth less than one who spends $250 at full price.
Step 3: Collapse scores into value tiers
Twenty-seven score combinations are too many to write content for. Group them into three value tiers:
- High value: recency 3 and frequency 3, or frequency 3 and monetary 3 with recency 2.
- Core: everyone with two or more orders who is not high value and has bought in the last 180 days.
- Low or lapsed: one-time buyers and anyone with recency 1.
Built-in RFM options
Shopify assigns each customer a built-in RFM group, with labels such as Champions, Loyal, At risk and Dormant. Klaviyo offers an RFM report that scores customers 1 to 3 on each dimension, but it requires a paid add-on (Advanced KDP) and a minimum order history. Either is a reasonable start; check how each defines its groups before building messaging on them.
What does the Buying Stage layer add?
The B in RFMB stands for Buying Stage: the furthest step a person reached in the last 30 days. It is measured from on-site behavior that your ESP already tracks when its Shopify integration and on-site script are installed.
| Stage | Event (Klaviyo naming) | What it signals |
|---|---|---|
| B1 Visit | Active on Site | Aware and returning, no product focus yet |
| B2 Product view | Viewed Product | Interest in specific items |
| B3 Add to cart | Added to Cart | A purchase decision is forming |
| B4 Checkout | Started Checkout | Ready to buy, something stopped them |
| B5 Purchase | Placed Order | Bought in the last 30 days |
Each person sits at their highest stage only. Someone who viewed products and then started checkout is B4, not B2. People with no tracked activity in 30 days have no stage, which is information in itself: their value score is the only thing you know about them.
On-site events only attach to known visitors (people who clicked an email, logged in or were identified another way), and the 30-day window rolls, so customers move back down as activity ages out.
How do value and intent combine into segments you can act on?
Value tells you how much you can afford to invest in a customer. Intent tells you whether now is the time. Put them in a grid and most of your messaging rules write themselves.
| Value tier | No stage or B1 | B2 to B3 (browsing, carting) | B4 (checkout, not bought) | B5 (bought, last 30 days) |
|---|---|---|---|---|
| High value | Early access, new arrivals, no offer | Personal product picks, restock alerts | Service reminder, no discount | Thank you, no promo for a while |
| Core | Regular campaigns at normal frequency | Browse follow-up with reviews | Cart reminder, free-shipping nudge if needed | Post-purchase flow toward next order |
| Low or lapsed | Reduced frequency, then a win-back campaign | Re-engagement with bestsellers | Smallest incentive that works, tested | Second-order push while interest is fresh |
Three rules come out of this grid:
- High value plus high intent means no discount. A loyal customer who started checkout is likely to buy. Remind and help them; don’t pay them to do what they were already doing.
- Low value plus high intent is where incentives belong. A lapsed one-time buyer who is carting again is the person a modest offer might actually move.
- B5 suppresses promotions. Anyone who bought in the last 30 days should leave discount campaigns automatically. Their next message should come from the post-purchase sequence, which drives your first-to-second order rate.
For a deeper look at why that second order matters so much, see The Second-Order Problem. For recovering lapsed customers, see our win-back flow playbook.
Worked example: what the grid saves (illustrative)
Illustrative numbers only. A store sends a 15% code to all checkout abandoners, about 600 people a month. Average order value is $120, so each redeemed code costs $18.
About 150 of those abandoners are high-value customers, and half of them (75) complete the order with the code, most of whom would have bought anyway. That is 75 × $18 = $1,350 a month, roughly $16,000 a year, from one flow. A no-discount reminder for high-value B4 customers recovers most of it; the abandoned cart playbook shows how to keep recovery rates up without the code.
How do you build RFMB segments in Klaviyo?
The same logic works in most ESPs. In Klaviyo:
- Store the value tier on the profile. Upload tiers as a custom profile property (for example
value_tier = high), or use built-in RFM groups. Refresh monthly. - Build the five stage segments. For B4: “Started Checkout at least once in the last 30 days” and “Placed Order zero times in the last 30 days.” Work down the stages the same way, so each person lands in one stage only.
- Create the combined segments you will actually use. Start with four or five: High value at B4, High value at B5, Low or lapsed at B3 or B4, Core with no stage, Low or lapsed with no stage.
- Wire them into flows and campaigns as flow filters, conditional splits (a no-discount branch for high value), campaign exclusions and suppression audiences synced to Meta and Google.
- Track revenue and margin per recipient by segment, not just opens and clicks.
Shopify Flow can also tag high-value customers in the store, so customer service and loyalty apps see the same tier.
The full build, with segment definitions and message templates, is in our RFM segmentation playbook.
What mistakes make RFM segments useless?
- Building 20 segments and writing content for three. A segment that receives the same email as everyone else is just a report.
- Scoring on gross revenue. Heavy returners end up in your VIP tier.
- Never refreshing the scores. RFM from six months ago describes customers who no longer exist in that state.
- Using opens to measure intent. Apple Mail Privacy Protection can register opens that never happened. On-site events and orders are far more reliable signals.
What should your first RFMB build include?
- [ ] A clean customer export scored 1 to 3 on recency, frequency and monetary value, with thresholds based on your repurchase cycle.
- [ ] Three value tiers stored as a profile property and refreshed monthly.
- [ ] Five Buying Stage segments on a rolling 30-day window, each person in one stage only.
- [ ] A no-discount branch for high-value customers in checkout and cart flows.
- [ ] Recent buyers (B5) excluded from promotional campaigns and paid retargeting.
- [ ] Low or lapsed customers on reduced frequency, with a win-back flow at the end.
- [ ] Revenue and margin per recipient reported by segment every month.
Once segments are live, check whether they move retention. The Retention Rate Calculator gives you a baseline to compare against each quarter.
FAQ
How often should I recalculate RFM scores?
Monthly works for most stores. Fast-moving consumables with short repurchase cycles can justify weekly updates, while high-ticket, low-frequency products can be refreshed quarterly. The Buying Stage layer updates continuously because it runs on live events.
Do I need a minimum number of customers for RFM segmentation?
The method works at any size, but segments need enough people to be worth separate content. With a few hundred customers, start with two value tiers and the B4 and B5 stages. Some built-in tools set their own minimums. Klaviyo’s RFM report, for example, asks for at least 500 customers who have ordered and 180 days of order history.
Should I use RFM or predicted customer lifetime value?
Both, if you have them. Predicted customer lifetime value estimates future spend; RFM explains past behavior in terms a team can act on. The Buying Stage layer adds timing, which neither provides.