RFM Segmentation for Shopify Stores: A Practical Guide

Three questions, in the order that matters.

1. Recency — how many days since the last order. On its own the strongest hint that someone will buy again.
2. Frequency — how many orders in total. It separates a habit from an accident.
3. Monetary — how much has been spent in total. It tells you what the habit is worth.

Recency only means something against your own reorder cycle. If you sell coffee, 60 days without an order is a customer drifting away. If you sell mattresses, 60 days is normal and so is 600. Work out the median gap between first and second order in your shop and read Recency against that number, not against a rule of thumb from a blog post.

The model itself is old and well tested — it was formalised for direct mail in the 1990s, in work that showed you get more profit from selecting a smaller, better list than from mailing everyone (Optimal Selection for Direct Mail, Marketing Science, 1995).

It scores each dimension from 1 to 5, then compresses the three digits into one named group.

Shopify's own RFM report gives every customer a three-digit score — one digit for Recency, one for Frequency, one for Monetary value, each between 1 and 5. A 5 means the customer sits in the top 20% of your store on that dimension and a 1 in the bottom 20%, which is why the result is relative to your shop rather than to a benchmark. The group is then decided by the Recency digit plus the average of the other two, rounded down (Shopify Help Center, retrieved August 2026).

That produces 11 groups, and the names carry the instruction:

• Champions — bought very recently, often, and spent the most.
• Loyal — recent, many orders, high spend.
• Active and Promising — recent buyers who have not built a habit yet.
• New — a very recent first purchase, low spend so far.
• Needs attention — recent, but middling on both other counts.
• At risk and Previously loyal — a strong history of orders and spend, and no recent order. The most expensive groups to ignore.
• Almost lost and Dormant — no recent order, and no strong history behind it either.
• Prospects — no orders yet at all.

Keaz shows its own RFM badge on every row of the contact list — Champions in green, Loyal in turquoise, and further groups beyond those. It is assigned automatically from purchase behaviour: derived from what a contact does, not something you type in or configure. Contacts: Overview shows where the badge sits and what the neighbouring columns mean.

Treat the two readings as separate. Shopify documents its own scoring and its own eleven groups; the badge in Keaz is Keaz's. Use whichever one you are actually going to act on instead of trying to reconcile them.

The two ends of the list, for opposite reasons — and almost never the middle.

1. At risk and Previously loyal first. These people proved they like your shop and then stopped. One well-timed message is the cheapest revenue in the account, because you are not buying a customer, you are reminding one. The same idea on a much shorter clock is winning back abandoned carts.
2. Champions and Loyal second, and not with a discount. They buy at full price already. Early access, a new drop, a genuine thank-you. A discount here is a margin loss on an order that was coming anyway.
3. New and Promising third, with something that helps rather than sells: how to use the thing they bought, what goes with it.
4. Dormant and Almost lost last, or not at all. A large group with a low response rate is the fastest way to make a WhatsApp channel look expensive.

Two mechanics do the work, and it pays to keep them apart. A Keaz flow is triggered by a Shopify event — an order, an abandoned checkout — not by a contact's RFM group. RFM is how you decide who is in an audience: build the group as a condition in the Segment Builder and it becomes a saved audience that newsletters and flows can send to. So a winback that starts from a purchase pattern rather than a single event is a segment plus a send, not a trigger.

Which one to do first is mostly a question of size rather than taste. How we calculate what a reachable customer is worth sets out the arithmetic, and the revenue forecast applies it to your own order data.

RFM tells you who is worth a message. It does not give you permission to send one. In Keaz those are two independent badges, and a Champion who is Unsubscribed is an ordinary row in the list, not an error.

Permission has its own four levels — Transactional, Double Opt-In, Single Opt-In and Unsubscribed — and a segment built on RFM alone will quietly drop everyone who does not clear the permission condition. That is almost always the answer when an audience looks large in the builder and small in the send.

So the order is: reach and permission first, groups afterwards. Most stores start out able to reach 35–55% of their last-12-months customers, and no segmentation model rescues a channel at 40%. Our guide to collecting WhatsApp opt-ins the GDPR-compliant way covers the principles, and consent stays an explicit, documented step of its own.

This is a summary rather than legal advice — have your own counsel review how you collect and document consent.

Not always. One threshold is easier to explain, easier to audit, and on a short list it often selects the same people.

RFM earns the extra complexity in three situations:

1. When recency and value disagree. A spend threshold cannot tell your best customer from your best former customer. That distinction is RFM's entire job.
2. When the list is large enough to split. Below a few hundred reachable contacts, eleven groups produce segments too small to learn anything from.
3. When you send often. Message monthly and you need a rule for who gets left out this time. “Spent over €200” is not one.

If none of the three apply, use the threshold and revisit in a quarter — simple beats sophisticated when nobody has time to maintain sophisticated. Reorder cycles also differ sharply by vertical, which changes where the line falls: what we see per industry is the shorter way to check yours.

Five patterns, in the order we run into them:

1. Treating RFM as permission. The most expensive one, and the reason the two badges are deliberately independent.
2. Discounting Champions. They were going to buy anyway, so the discount is a straight margin loss wearing a campaign's clothes.
3. Reading Recency against a generic rule. 90 days means something for coffee and nothing for furniture. Use your own median reorder gap.
4. Segmenting before you have reach. Eleven groups across a 40% reachable base leaves you with slivers. Coverage first, then groups.
5. Setting it up once. Groups move as behaviour changes, so a segment built in March describes different people in June. That is the point — and it means the message has to still make sense for whoever is in there now.

Can I change the RFM thresholds in Keaz?

The badge is assigned automatically from purchase behaviour and is derived rather than entered. If you need a boundary specific to your shop — a spend level, an order count, a date — build it as a condition in the Segment Builder instead.

How many groups should I actually message?

Two or three. Pick At risk, Champions, and one growth group, and give each a different message. More than that and nobody keeps them current.

Do I need an email tool for this?

No. RFM comes out of Shopify purchase data, so a connected Shopify store is enough. Connecting an email tool adds a second coverage metric, not the segmentation.

What about customers with no orders?

Shopify files them as Prospects and they carry no RFM signal at all. They are an opt-in problem, not a segmentation problem.

Does a segment update itself?

Yes — that is the reason to save it rather than export a list. The condition stays, the people in it change as they buy or stop buying.