segmentation
Installation
SKILL.md
When to segment
Aggregate metrics often hide divergent behavior across subgroups. Segment before drawing conclusions when:
- The user asks "who are our best/worst X" or "how does X vary across groups"
- A trend looks flat but you suspect offsetting movements (e.g. SMB churn masked by enterprise growth)
- You're sizing or prioritizing — different segments deserve different strategies
Common segmentation types
- RFM (Recency, Frequency, Monetary) — retention and marketing on transactional businesses
- Behavioral — by feature usage, lifecycle stage, engagement level
- Value-based — ARR tier, deal size, plan
- Demographic / firmographic — for B2C / B2B targeting
- Acquisition — by channel, campaign, signup cohort
How to evaluate a segmentation
- Size — each segment should be material (rule of thumb: ≥5% of base, or large enough for statistical confidence)
- Distinctness — segments should differ on the metric you care about, not just on the segmenting variable
- Stability — membership shouldn't churn so fast the segment becomes meaningless