Win-Back (Lapsed Customer Re-engagement)
The signal. Unlike churn prevention (where signals are probabilistic), win-back triggers on a hard event: cancellation confirmed, subscription lapsed, or X days since last purchase/login. The signal is deterministic and usually available in transactional systems — no propensity model required to identify the audience.
Why it's distinct from churn prevention. Churn prevention intervenes before the decision is made. Win-back intervenes after. The intervention window, offer type, and channel mix differ: win-back campaigns must balance enough incentive to re-engage against training customers to churn-and-return cyclically to extract offers.
Latency pattern. Win-back is typically a batch-to-intra-day use case. The trigger is a cancellation or lapse event; the response is a sequenced campaign over days or weeks — Day 1, Day 7, Day 14, Day 30 touchpoints. Unlike churn prevention (which needs fast activation from a live risk signal), win-back sequences are orchestrated in advance.
Architectural implications.
- Model output (propensity to re-engage by segment) computed at Daily latency; activations at Intra-day or Daily latency for each sequence step.
- Multi-touch sequence requires a campaign orchestration layer, not just a single activation.
- Suppression: customers who re-engage must immediately exit the win-back sequence. Requires real-time feedback from the activation system back to the segment definition — a composable-stack coordination challenge if the sequence lives in an ESP and the segment lives in the CDW.
Relevant dimensions.
- Marketing goal: org-dim.marketing-goal.retention, org-dim.marketing-goal.customer-lifetime-value.
- Industries: org-dim.industry.telecommunications (lapsed subscribers), org-dim.industry.ecommerce (lapsed buyers), org-dim.industry.retail (loyalty program dormants).