A segmentation model is the recipe for how you split your customer base into groups. Different models suit different goals — predicting churn, prioritizing acquisition, designing lifecycle programs — and the choice of model shapes which actions become possible.
Five segmentation models worth knowing
- RFM (Recency, Frequency, Monetary) — Score each customer on how recently they bought, how often, and how much. Classic ecommerce segmentation; works well for non-subscription orders.
- Lifecycle stage — New, active, at-risk, lapsed, reactivated. Maps directly to subscription operations; the most-used model for retention.
- Behavioral clustering — Use unsupervised learning (k-means, hierarchical clustering) to find natural groups in behavior data. Surfaces segments you might not have hypothesized.
- Value-based — High-LTV, medium, low. Allocates retention investment in proportion to expected return.
- Needs-based — Group by what the customer is trying to achieve (convenience, novelty, identity, anxiety reduction). Often qualitative; aligns with Jobs-to-be-Done.
Which model fits subscription businesses
For most subscription operations, lifecycle stage + value-based is the highest-leverage pairing. Lifecycle stage tells you which message to send (onboarding, save, loyalty, win-back); value-based tells you how much to spend on the message. A long-tenured high-LTV subscriber warrants a personal email; a 30-day-old low-AOV subscriber gets the automated sequence.
Implementation
- Start with lifecycle. The five segments — new, active, at-risk, lapsed, reactivated — are operationally clear and easy to tag.
- Layer value on top. Within each lifecycle segment, identify the top 20% by LTV — they get personal touch.
- Add behavioral signals over time. Skip patterns, engagement, portal use refine the segments further.
- Avoid over-engineering. A simple model used consistently beats a sophisticated one used inconsistently. Pick your level and ship.