How to Segment Customers in Ecommerce: A Step-by-Step Method

Segmenting customers means splitting them into groups that should be treated differently, then treating them differently. For an online store, the useful groups come from orders, checkout and subscription data, not from personas. This guide gives a six-step method, the main segmentation types, RFM with a worked example, the payment segments most guides skip, and how to build all of it in Shopify and Klaviyo.

What is customer segmentation?

Customer segmentation is dividing your customers into groups that share a trait or a behavior, so each group can get its own message, offer or exclusion. "Bought twice, nothing in 90 days" is a segment. So is "renewal failed this week" and "lives in Canada".

In current tools, a segment is a rule, not a list. Shopify describes its customer segments as "dynamic, rule-based customer lists": customers are added or removed automatically as they meet the criteria or stop meeting them.1Shopify Help Center, Customer segmentation Klaviyo says the same of its segments, which update "in close to real time", while its lists only grow when people subscribe or are added.5Klaviyo Help Center, Getting started with segments That matters: a CSV exported last month doesn't know who ordered, refunded or disputed since.

How to segment customers in 6 steps

The method below works for a store with a few hundred customers and for one with a few hundred thousand. Only the tooling changes.

Step 1: start from a business question

The question decides which data you need and which method fits. Shopify's own guide to ecommerce segmentation recommends the same order: goal first, then data, then segments. The table pairs common store questions with a first segment and the metric that will judge it.

Step 2: the data you need

Most store segments need four sources: the store or order system, the checkout, the subscription billing system, and the payment processor. Email and SMS engagement come from the messaging tool. What the customer told you (quiz answers, preferences) is zero-party data and usually lives in a form or survey tool.

Before building on these fields, check three things:

• One customer, one record. If the same person exists under two emails in the store and in the billing tool, their order count and recency are wrong everywhere. How records link together is covered in what CRM databases are.
• Shared definitions. "Active subscriber", "churned" and "repeat buyer" must mean the same thing in every tool. Otherwise two segments with the same name hold different people.
• Freshness. A payment segment that syncs once a day will email a customer about a failed renewal hours after they fixed it.

Types of customer segmentation

Most guides list four classic types: demographic, geographic, psychographic and behavioral. Ecommerce adds value-based segmentation, and stores with subscriptions need a sixth, built on payment and billing state.

Behavioral segments get their own guide: what behavioral targeting is covers the signals, first-party data and the legal limits on profiling, including for health-adjacent products. Value-based and payment segments are where most stores get the fastest results, so the next two sections go deeper.

RFM segmentation explained

RFM is the standard way to segment by value using only order data. Every store already has the three inputs.

Each dimension gets a score from 1 to 5, relative to your own customers. Shopify's method is a good reference because it's documented: a score of 5 means the customer is in the top 20% of that dimension for your store. Shopify then averages frequency and monetary into one FM score, rounded down, and places each customer in a group from R and FM.4Shopify Help Center, Customers reports (RFM customer analysis)

A worked RFM example

Take ten customers of a store, scored on October 1. With ten customers, each quintile holds two: the two most recent buyers get R = 5, the next two R = 4, and so on. Same for orders (F) and spend (M).

Applied to the ten customers, and with what to do for each group:

The "what to do" column follows the goals Shopify gives for each group (move Promising and New to Active, Loyal to Champions, keep Champions as advocates).4Shopify Help Center, Customers reports (RFM customer analysis) The point of RFM is visible in G and H: by recency alone they look lost, but their history says they're the customers most worth a personal message.

RFM for subscription brands

RFM was built for repeat one-time purchases. With subscriptions, every renewal counts as an order, so a subscriber who never chose to buy again still scores high on frequency. Two fixes: compute RFM on one-time orders only and segment subscribers by subscription state instead, or keep RFM and always read it next to subscription status. A subscriber in "Champions" whose renewal just failed belongs in the payment segment below first. Lifetime value by customer is covered in what CLTV is.

Payment and subscription segments

Marketing guides rarely mention them, but billing and payment data produce some of the most important segments a store has. Most of them change what you should not send.

Stripe's subscription statuses are a good map. past_due means payment on the latest invoice failed or wasn't attempted; Stripe may retry while the subscription is in that state. unpaid means retries are over but the subscription still exists, and Stripe says to revoke access at that point. canceled is final.6Stripe Docs, How subscriptions work (subscription statuses) Your billing tool may use other names, but the states are the same.

The open-dispute row is a hard rule, not a preference. Stripe states that you can't issue a refund outside the dispute process while the dispute is open.7Stripe Docs, How disputes work A promotional email to that customer for the disputed product helps nobody. The retry and email sequence for failed renewals is covered in dunning and failed payment recovery and subscription payment failures; how to keep disputes from happening is in chargeback prevention.

Inside Tagada

Build the funnel once, route every customer to the right offer

Checkout, upsells and subscriptions on one flow, with conditions per country, order or customer.

Start free

Segments vs cohorts

A cohort is a group defined by when customers started: everyone whose first order was in March, everyone who subscribed during a launch. A segment is defined by what is true about customers now. Both are useful; they answer different questions.

The two work together: a cohort table shows that customers acquired through one campaign stop buying after the second order, and a segment then targets that group before the third. How to build and read a retention table is in what cohort analysis is.

Step 4: check the segment size before you act

A segment you can't measure is a guess. The number of customers you need depends on how often the outcome happens and how small a difference you want to detect. The chart shows how many customers you need in each group (targeted and holdout) to detect a 20 to 25% relative change in the conversion rate.

What this means in practice:

• Rare outcomes need big segments. A win-back email that turns 2% of lapsed customers into buyers needs tens of thousands of customers in total to be measured. Most small stores can't test it per segment.
• Merge segments that would get the same action. "At risk" and "Previously loyal" can share one win-back campaign and one holdout.
• Widen the window instead of shrinking the segment. Run the test for longer, or measure a more frequent outcome (any order within 60 days rather than a specific product).
• Some segments don't need a test. Exclusions (open dispute, unsubscribed) and service messages (failed renewal) are done because they're right, not because they lift revenue.

Steps 5 and 6: act, then measure with a holdout

Each segment gets one action and one holdout. The holdout is a random share of the segment that gets nothing. It's the only way to know what the campaign changed, because many customers in a segment would have bought anyway.

Three rules make the comparison honest:

1. Split at random, before sending. Not "people who opened" versus "people who didn't".
2. Compare orders and margin, not opens or clicks. Margin matters most for discount campaigns: a coupon that goes to people who would have bought at full price loses money while showing great conversion.
3. Measure over the same period. Seasonality makes "this month vs last month" useless.

Where the action runs (email, SMS, on-site, ad audiences built from customer lists) and how automations use these segments is covered in commerce marketing automation.

How to build customer segments in Shopify

Shopify builds segments in the customer segment editor with a query language: you combine filter names, operators and values into a query.1Shopify Help Center, Customer segmentation Several filters are joined with AND (the customer must meet all of them) or OR (one is enough). Without brackets, AND is evaluated before OR, and Shopify recommends grouping filters with brackets.3Shopify Help Center, Components of a customer segment

-- At risk customers you can email
rfm_group = 'AT_RISK' AND email_subscription_status = 'SUBSCRIBED'

-- First-time buyers from the last 30 days
number_of_orders = 1 AND first_order_date >= -30d

-- Big spenders who have gone quiet
amount_spent > 500 AND last_order_date < -90d

-- Abandoned a checkout this week
abandoned_checkout_date >= -7d

-- Bought a tagged product in the last 3 months
products_purchased MATCHES (tag = 'refill', date BETWEEN -3m AND today)

-- Exclusion list kept with a customer tag
customer_tags CONTAINS 'dispute_open'

The filters worth knowing, all from Shopify's reference:2Shopify Help Center, Shopify-based customer segment filters

• rfm_group takes Shopify's own RFM groups: 'CHAMPIONS', 'LOYAL', 'ACTIVE', 'NEW', 'PROMISING', 'NEEDS_ATTENTION', 'ALMOST_LOST', 'AT_RISK', 'PREVIOUSLY_LOYAL', 'DORMANT', 'PROSPECTS'. You get the RFM model without computing it.
• predicted_spend_tier ('HIGH', 'MEDIUM', 'LOW') is only available once the store has made more than 100 sales.
• email_subscription_status and sms_subscription_status keep consent in the rule itself. Put them in every marketing segment.
• customer_tags is how data from other tools gets in. If your subscription app, helpdesk or payment tool tags customers (payment issue, dispute open), you can segment and exclude on that tag. Tags aren't case sensitive.
• customer_countries uses two-letter ISO codes, for example customer_countries CONTAINS 'US'.

Shopify segments only know what Shopify knows. A failed renewal handled by a separate billing tool won't appear unless that tool writes it back as a tag or a metafield.

How to build customer segments in Klaviyo

Klaviyo segments are built from conditions, the most used being "What someone has done (or not done)", for example Placed Order a number of times over a period, and "Properties about someone".5Klaviyo Help Center, Getting started with segments Four details from Klaviyo's own help change how you build them:

1. The default operator between conditions is OR. Switch to AND when every condition must be true. AND makes the segment stricter, OR more inclusive.
2. OR with a negative condition can catch almost everyone, because each condition is checked separately. "Has not placed an order in 90 days OR has not opened an email" includes nearly the whole list.
3. Segments don't require opt-in. A segment can contain people who never consented to marketing, so add a consent condition before sending a campaign to it.
4. Snapshots freeze a segment. A snapshot creates a static list of the members at that moment; the segment itself keeps updating. Useful to keep the exact audience and holdout of a past test.5Klaviyo Help Center, Getting started with segments

Only top-level event properties can be used, and property values are case sensitive, so "payment_failed" and "Payment_Failed" are two different values.5Klaviyo Help Center, Getting started with segments Large segments can also take a while to populate, so build them before the day you plan to send.

Common customer segmentation mistakes

Where Tagada fits

Whatever stack you use, the test is the same: can the tool that holds your orders, renewals, refunds and disputes pass those events to the place where you build segments, fast enough for a failed renewal to reach the right customer before the next retry?