Overview
Rather than new acquisition, retaining existing customers is cheaper and more profitable— this skill designs the retention that marketing can do, based on that premise. It catches early churn signals to intervene in advance, holds onto customers at the moment they try to leave with save offers, brings back already-departed customers with winback sequences, and maintains relationships with lifecycle-stage messages.
This skill focuses on messages, campaigns, and metrics aspects. It does not cover purchase-cycle calculation, subscription-plan operations, or unit-economics modeling, and delegates those to the moai-seller commerce skill (see the responsibility boundaries below).
Reference source: github.com/coreyhaines31/marketingskills (MIT — full notice: repository root NOTICE) — the methodology of the churn-prevention skill has been restructured from a marketing-lifecycle perspective.
Trigger keywords
retention, customer retention, churn prevention, churn rate, churning, winback, reactivation, re-engagement, lifecycle messages, health score, cancellation flow, save offer, retention curve, cohort, re-visit inducement
Practical retention patterns
1. Churn Classification (two types)
Churn requires different responses depending on its cause.
| Type | Share | Cause | Marketing response |
|---|---|---|---|
| Voluntary churn(Voluntary) | 50-70% | Customer cancels directly | Cancellation flow, churn survey, save offer |
| Involuntary churn(Involuntary) | 30-50% | Payment failure, card expiration | Dunning sequence, advance notification |
Marketing mainly handles the lifecycle messages for voluntary churn. Since payment-retry logic and dunning operations for involuntary churn overlap with the commerce and subscription-operations domain, this skill supports only dunning email copy, and delegates retry policies.
2. Early churn signals & health score
Churn signals appear 2-6 weeks before cancellation. Rather than post-hoc winback, advance interventionhas a higher ROI.
Leading indicators:
- 50%+ decrease in login and visit frequency
- Stopping use of core features and core behaviors
- Changes in support-inquiry patterns
- Increased visits to pricing and cancellation pages
- Starting data export (a strong churn signal)
Health score(weighted sum of engagement, usage, support sentiment, and payment health):
| Score | Grade | Response |
|---|---|---|
| 80-100 | Healthy | Advocacy·referral inducement |
| 60-79 | Attention needed | Value re-emphasis content |
| 40-59 | Risk | Re-engagement campaign |
| 0-39 | Severe | Immediate intervention·save offer |
3. Cancel Flow Architecture (Cancel Flow)
The last touchpoint at the moment of leaving. Even with only a basic cancel flow, 10-15% save possible.
트리거 → 이탈 설문 → 동적 세이브 오퍼 → 확인 → 취소 후(윈백 트리거)
Exit survey(5-8 reasons): price burden · low usage · lack of features · competitor migration · technical issues · temporarily unnecessary · business closure.
Dynamic save offer(matched to reason):
| Reason | Save offer |
|---|---|
| Price burden | 20-30% discount for 2-3 months or plan downgrade |
| Low usage | 1-3 month pause or onboarding re-support |
| Lack of features | Roadmap preview + expected timing |
| Competitor migration | Comparison data + incentive |
| Technical issues | Immediate support escalation |
Key: Instead of inducing guilt ("Are you really leaving?"), value re-emphasis + tailored offer. Even after cancellation, make the easy reactivation path prominent.
4. Winback·Re-engagement Sequence (Winback / Re-engagement)
Re-engagement(still a customer, low activity) vs Winback(already churned) are different.
Re-engagement sequence(health score decline trigger):
| Timing | Purpose |
|---|---|
| Signal detection D+0 | Value reminder (missed key features·benefits) |
| D+3 | Success cases·social proof |
| D+7 | Personalized incentive or 1:1 support offer |
Winback sequence(triggered immediately after cancellation):
| Timing | Purpose |
|---|---|
| Cancellation D+0 | Thanks + easy reactivation link |
| D+7 | "What has changed" (new features·improvements) |
| D+30 | Strong comeback incentive (limited) |
| D+90 | Final offer or wrap-up |
5. Lifecycle Messaging (Lifecycle Messaging)
Automate messages suited to each stage—awareness→activation→habituation→loyalty→churn risk—on a trigger basis. For triggers such as a sharp drop in usage, reaching a feature limit, or support ticket escalation, respond proactively with value reinforcement content(guilt messages prohibited).
6. Retention Metrics Framework
| Metric | Definition | Reference target |
|---|---|---|
| Monthly churn rate | Churned customers / customers at start of month | B2C < 5% |
| Revenue churn rate | (churned MRR − expansion MRR) / MRR at start of month | Aim for net negative |
| Cancellation save rate | Saves / Total cancellation sessions | 25-35% |
| Retention curve | Survival rate over time by cohort | Flattening-oriented |
Cohort segmentation: Break down churn patterns by acquisition source, pricing plan, tenure, and cancellation reason to find structural vulnerabilities. If the retention curve drops sharply early on and then flattens, it's healthy; if it keeps declining, it's a PMF/value delivery problem.
Retention anti-patterns
- Relying only on post-hoc winback, no early signals or preemptive intervention (more expensive)
- Guilt-tripping/coercive messaging ("Are you really going to leave?")
- No save offer in the cancellation flow (losing a 10-15% save opportunity)
- Treating re-engagement and winback as the same sequence (different states)
- In promotional re-engagement emails,
(광고)·missing opt-out (violation of the Information and Communications Network Act) - Looking only at overall churn rate without retention curves or cohorts (can't find the problem segment)
Workflow
Step 1: Confirm retention goals and targets
Confirm the improvement metrics (churn rate/save rate/reactivation rate), churn type (voluntary/involuntary), target segments, and whether lifecycle automation currently exists.
Step 2: Churn diagnosis
Organize the proportion of churn types, early signal definitions, health score design, and churn patterns by cohort (retention curves) to decide where to intervene.
Step 3: Design intervention points
Prioritize among preemptive intervention (health score triggers), the cancellation moment (cancellation flow + dynamic save offer), and post-hoc (winback sequence), and define the triggers and conditions for each.
Step 4: Write messages and sequences
Draft the purpose, subject, and body for each stage of the re-engagement and winback sequences, the save offer copy, and lifecycle messages. If it's a promotional send, the requirements of the Information and Communications Network Act §50(subject (광고)·sender information·free opt-out·nighttime sending restrictions) must be reflected.
Step 5: Design metrics and measurement
Set targets for churn rate, save rate, reactivation rate, and retention curve, and define the A/B plan for save offer types and copy and tracking of LTV from accepted offers.
Usage examples
Example 1: "Subscription churn is severe. Add a save offer to the cancellation flow" → 5-8 cancellation survey questions → dynamic offers by reason → confirmation/reactivation path → save rate target
Example 2: "Build a winback email sequence for customers inactive for 90 days" → winback trigger → D+0/7/30/90 sequence → comeback incentive → Information and Communications Network Act check
Example 3: "Create a health score and preemptive intervention campaign using early churn signals" → define leading indicators → 4-tier score → re-engagement messages by tier → trigger automation
Output format
- Retention diagnosis summary (proportion of churn types · early signals · retention curve interpretation)
- Health score design (indicator weights · responses by tier)
- Cancellation flow + dynamic save offer mapping
- Re-engagement / win-back sequence (purpose, subject, body draft by stage)
- Retention metric KPI design (churn rate, save rate, cohort)
- Information and Communications Network Act compliance checklist (for advertising messages)
Precautions
| Situation | Response |
|---|---|
| When there is no customer behavior/churn data | Build a general sequence based on industry benchmarks, then recommend segmentation once real data is secured |
| When the distinction between advertising/informational messages is ambiguous | Design conservatively based on advertising criteria (Information and Communications Network Act §50), then recommend legal review |
| Requests for repurchase cycle, subscription pricing plans, or unit economics calculations | Outside the scope of this skill. Refer to the moai-seller skill under the responsibility boundaries below |
| Requests for payment retry policy for involuntary churn | Support only dunning copy; retry logic is delegated to the subscription operations domain |
Post-processing (required chain — text outputs only)
Re-engagement/win-back email body and subject, save offer copy, lifecycle messages, etc. Text outputsalways go through the post-processing chain before sending. Structural/numerical outputs such as health score tables, retention curves, and churn rate KPIs are not subject to this.
Recommended chain (text): marketing-retention(텍스트) → moai-coworker:ai-slop-reviewer → moai-writer:korean-humanize → 최종 검수
Related skill
| Skill | When to use |
|---|---|
moai-marketer:content-email-sequence |
Produce win-back/re-engagement as actual email automation sequences (including legal requirements) |
moai-marketer:marketing-campaign-planner |
Expand retention into a full campaign/CRM strategy |
moai-marketer:marketing-performance-report |
Retention metrics/cohort performance report |
moai-seller:commerce-repurchase-timer |
Purchase-cycle-based repurchase timing (commerce purchase timing prediction) |
moai-seller:commerce-subscription-strategist |
Subscription operations, pricing, plan design |
moai-seller:commerce-ltv-cac-architect |
Unit economics (LTV/CAC) modeling |
moai-coworker:ai-slop-reviewer |
Post-chaining (required, text): AI pattern review of retention messages |
moai-writer:korean-humanize |
Post-chaining (required, text): next step after ai-slop-reviewer to remove Korean AI tell-tale signs |
When not to use this skill
- Purchase-cycle-based repurchase timing prediction:
moai-seller:commerce-repurchase-timeruse - Subscription pricing plan/plan/subscription operations design:
moai-seller:commerce-subscription-strategistuse - Unit economics (LTV/CAC) calculation/modeling:
moai-seller:commerce-ltv-cac-architectuse - Actual email automation sequence production (including legal requirements):
moai-marketer:content-email-sequenceuse