Abstract talk about segmentation rarely translates into action. Concrete examples — the segments other subscription merchants actually use — make the practice usable. Below are the segments that show up in most healthy subscription operations, with the criteria and the action each unlocks.
Five high-leverage subscription segments
- New subscribers (0–60 days)
Criteria: signed up within the last 60 days, fewer than 3 cycles completed.
Action: heavy onboarding emails, first-skip education, early-life check-in survey. - At-risk (recent skip + support contact)
Criteria: skipped 2+ consecutive cycles, opened a support ticket in the last 30 days, or engagement dropped 50%+.
Action: personal email, save offer, cadence review. - Long-tenured loyalists (12+ months)
Criteria: 12 months of continuous active subscription, no recent failed payments.
Action: surprise gift, referral ask, exclusive product access. - Heavy users (top 20% by AOV or frequency)
Criteria: AOV in the top quintile, or above-average cycle frequency.
Action: expansion offer (additional product line), early access to launches. - Lapsed within 90 days
Criteria: cancelled in the last 90 days, no current active subscription.
Action: win-back campaign with a tailored offer based on cancellation reason.
Examples by product category
- Coffee subscription — Variety-seekers (frequent swaps) vs. loyalists (same blend every time). Different cadence and product messaging for each.
- Vitamin subscription — High-consumption (90-day cycle) vs. lighter users (180-day cycle). Default cadence should match observed consumption.
- Beauty box — Discovery-stage (new, exploring) vs. preference-set (knows what they like). Personalize the box around the segment.
- Pet food — Single-pet vs. multi-pet households. Different plan tier defaults and AOV.
How to build your own segments
Start with the actions you want to run differently (different email, different offer, different cadence), then define the criteria that identify the customers those actions are for. Backwards from the action is faster than forwards from the data — it keeps you out of segments that are real but useless.