Customer segmentation models compared

Start with rules over purchase history when you need customer groups that your team can explain and act on. Use RFM to organize buying activity, value models to prioritize spending, and clustering to explore patterns you haven't defined. Predictive models suit a specific future outcome, provided you can check their accuracy on later purchases.

Customer segmentation models serve different decisions. Choose the model that changes your next message, offer, or service decision with data you can trust.

What is a customer segmentation model?

A customer segmentation model is a method for assigning customers to groups that receive different treatment. A segment is the resulting audience, such as buyers who ordered once and haven't returned.

Two choices often get mixed together: which customer traits matter, and how you form the groups. Location, spending, and product needs are traits. Rules, scoring, and clustering are ways to group customers using those traits.

A clear model helps your team agree on who belongs in an audience and why the message should differ. That can support more relevant service and less wasted promotion, but better results still need testing. For the wider setup process, see our guide to customer segmentation in Shopify.

Which model fits your decision?

Use the simplest model that separates customers who need different actions. The main choices differ in what they answer and how much work they require.

Rules suit a known action. A hypothetical skincare store could select customers who bought cleanser but haven't bought moisturizer. It could send advice about using the products together, while excluding customers with an unresolved complaint.

That is behavioral segmentation: the rule uses what a customer did. It is easy to explain and inspect. Its weakness is the assumption behind the rule: buying cleanser doesn't establish that someone needs another product.

RFM suits a broad view of buying activity. Recency means how recently someone ordered, frequency means how often, and monetary value means how much they spent. These signals can separate recent newcomers, regular buyers, and customers whose activity has slowed.

RFM and value models overlap, but they answer different questions. A customer with a large old purchase can rank highly for spending while showing weak recent activity. Someone who buys small orders regularly may deserve a different message from that customer.

Set activity labels around your products and buying patterns. A furniture customer can be satisfied without ordering again soon; a consumables buyer may have more frequent needs. Our study of customer segments and the next order explores why timing matters.

Value models suit decisions about attention and cost. Historical spend can identify customers worth recognizing with service or early access. If the decision involves a costly benefit, examine refunds and the cost of serving the customer too. A high revenue customer can still be expensive to retain.

Historical spending also differs from predicted lifetime value. One describes purchases already made; the other estimates what may happen later. Label them clearly before setting a service budget.

How much does a top-spender model cover?

A top-spender audience can matter without representing most of your business. Across 1,152 Shopify stores, the median share of annual revenue from the highest-spending tenth of customers was 35.5%. The period was September 2025 through August 2026.

Your own revenue concentration may differ. A VIP program also needs a benefit that fits the customer and a cost the business can support. This benchmark doesn't establish future spending or show that a VIP campaign causes extra purchases.

When do customer traits help?

Use a trait when it changes the customer's needs or the experience you can provide. These familiar segmentation types can work through straightforward rules.

Geographic segmentation uses location. A store could promote local pickup to nearby customers or separate shipping messages by destination. Check the address is relevant to the buyer; a gift delivery may go elsewhere.

Demographic segmentation uses attributes such as age or life stage. These can inform product research when customers provide them voluntarily. Avoid guessing someone's identity or needs from a name or a single order.

Psychographic segmentation uses interests, values, and lifestyle. A customer survey might distinguish buyers who care most about repairability from those focused on appearance. Use their answers to shape content, and leave unanswered questions unknown.

Needs segmentation focuses on the problem the customer wants solved. For example, a bag buyer might need daily commuting space or lightweight travel storage. Ask about intended use rather than assuming the purchased product tells the whole story.

Technographic segmentation uses devices or software. It can help diagnose a checkout experience that differs between phones and desktops. Device type alone is a weak reason to send different product offers.

Firmographic segmentation uses company traits, such as industry or business size. It is relevant when your store serves wholesale accounts whose ordering and service needs differ from retail buyers.

These groups depend on reliable records. Keep survey answers connected to the correct customer, distinguish purchases from browsing, and check missing fields before choosing a model. A detailed model built on guesses gives your team more labels without better decisions.

When should you use clustering or predictions?

Use clustering when you want to discover groups rather than write their membership rules in advance. It searches for customers with similar combinations of traits, such as basket contents, spending, and buying activity.

Methods such as k-means cluster analysis form groups from similarities in the supplied data. The groups still need interpretation and a different action. If two clusters would receive the same message, keeping them separate may add work without helping customers. A cluster driven mainly by an unusually large wholesale order may tell you little about retail customers.

Prediction starts with an outcome, such as whether someone will order again during a stated period. Check the model against later outcomes it didn't learn from, and compare it with a straightforward rule. Choose it when the improvement supports a better decision and someone can maintain it.

A cohort answers another question: how do customers who joined in the same period behave as time passes? It helps compare buying patterns across acquisition periods. It doesn't automatically identify who should receive today's message.

For a store with limited purchase history or little time to maintain models, clear rules are a practical starting point. Add complexity when it answers a question those rules leave unresolved.

Combine models and test the action

Combine models when the overlap changes treatment. In a hypothetical outdoor store, spending could identify valued customers, while a stated hiking interest selects useful product advice. Recent purchasers could receive care instructions before another sales offer.

Keep overlapping audiences from receiving conflicting messages. Agree which message takes priority, and check permission to contact customers in the channel you plan to use. A customer's membership in a segment doesn't establish that permission.

By the Numbers supports customer groups built from order history, spending, and behavior. Whatever tool you use, inspect sample members before activating the group and assign someone to maintain its definition.

Test the action as well as the model. Hold a random part of the eligible audience out of the campaign, then compare purchasing over the same period. Review contribution after offer and delivery costs, alongside complaints and unsubscribes. Customers who already buy more aren't proof that your campaign helped.

Document the rule, data source, owner, and intended action. Revisit it when products or buying patterns change. Merge groups that receive the same treatment, and retire groups nobody uses. Our customer segmentation strategy guide covers turning that choice into a campaign plan.

How we measured it

These benchmarks use anonymized, aggregated data from Shopify stores in the By the Numbers network. No individual store is identifiable from any figure. Stores that use an analytics app are not a random sample of all Shopify stores, so read the figures as a reference point, not as an average for the whole market.

Keep reading

If you want the full picture on customer segmentation, start with the customer segmentation playbook.

Related reading: How to build a customer segmentation strategy, Behavioral Segmentation: Store Examples and Which Customer Segments Predict the Next Order?.

By the Numbers builds this into the dashboard. See customer segments.