kai-retention — Customer Retention System
Design a complete retention system: churn diagnostics, retention tactics, engagement scoring, loyalty mechanics, and win-back campaigns.
Phase 0: Load Product Context
Check if MARKETING.md exists in the project root (same directory as CLAUDE.md, README.md, package.json).
If it exists: Read it — skip product discovery questions. It has the product name, ICP, value prop, monetization, brand voice, current channels, and competitive landscape.
If it does NOT exist: Auto-explore the codebase to create it in the project root (next to CLAUDE.md). Do NOT ask the user what the product is. Read CLAUDE.md, README.md, PROJECT.md, package.json, landing pages, and any project files. Search for email/ad/analytics config. Then create MARKETING.md using the template from /kai-email-system. Present draft to user for confirmation.
References
Load these files as context before starting:
E:\Dev2\kai-cmo-harness-work\knowledge\playbooks\customer-retention.mdE:\Dev2\kai-cmo-harness-work\knowledge\playbooks\growth-loops-applied.mdE:\Dev2\kai-cmo-harness-work\knowledge\channels\email-lifecycle.mdE:\Dev2\kai-cmo-harness-work\knowledge\personas\_persona-index.md
Phase 1 — Discovery
- Read from
MARKETING.md. Only ask about things not covered there:
- Business model (SaaS, ecommerce, services, marketplace)
- Current churn rate (monthly/annual, if known)
- Customer count and average revenue per customer
- Current retention efforts (any emails, loyalty program, support)
- Known churn reasons (from exit surveys, support tickets, cancellation flow)
- Product usage data availability (do they track feature adoption?)
- Customer segments (free vs. paid, plan tiers, cohorts)
- Identify the retention maturity level:
- Level 0: No retention effort beyond the product itself
- Level 1: Basic cancellation flow + occasional check-in emails
- Level 2: Lifecycle emails + usage tracking + support triggers
- Level 3: Predictive churn scoring + proactive intervention + loyalty program
Phase 2 — Analysis
Churn Diagnostics
- Categorize churn types:
- Voluntary: Customer actively cancels (dissatisfaction, budget, switched)
- Involuntary: Payment failure, expired card, billing issue
- Passive: Stops using but doesn't cancel (ghost users)
- Map the churn timeline: when do most customers leave?
- First 30 days (onboarding failure)
- 60-90 days (value not realized)
- At renewal (annual plan decision point)
- After price increase or feature change
- Identify leading indicators of churn:
- Login frequency decline
- Feature usage drop
- Support ticket volume spike
- NPS/CSAT score decline
- Billing page visits
Engagement Scoring Model
Define a health score (0-100) based on:
| Signal |
Weight |
Scoring |
| Login frequency (last 14 days) |
25% |
Daily=100, Weekly=60, Monthly=20, None=0 |
| Core feature usage |
25% |
Used all=100, Used some=50, Used none=0 |
| Support interactions |
15% |
Positive=80, Neutral=50, Negative=20 |
| Account expansion signals |
15% |
Upgraded=100, Stable=50, Downgraded=10 |
| Referral/advocacy |
10% |
Referred=100, NPS promoter=60, Passive=30 |
| Billing health |
10% |
Current=100, Late=30, Failed=0 |
Risk tiers: Green (70-100), Yellow (40-69), Red (0-39).
Phase 3 — Produce
Build these deliverables:
Retention Playbook
Intervention strategies by risk tier:
Red (0-39) — Immediate Rescue
- Trigger personal outreach within 24 hours
- Offer concession (discount, extended trial, premium support)
- Escalate to customer success manager
- Deploy win-back email sequence
Yellow (40-69) — Proactive Nurture
- Send usage tips targeting unused features
- Invite to office hours or webinar
- Share relevant case study or success story
- Request feedback (short survey, not NPS)
Green (70-100) — Expansion & Advocacy
- Request referral or testimonial
- Offer early access to new features
- Invite to advisory board or beta program
- Cross-sell or upsell relevant add-ons
Win-Back Campaign
For customers who have already churned:
- 3-email sequence: Day 1, Day 7, Day 30
- Each email addresses a different churn reason
- Include a specific offer or product update
- Run through quality gates before sending
Loyalty Program Design (if applicable)
- Reward mechanics: points, tiers, milestones, or referral credits
- Earning actions mapped to business goals
- Redemption options that drive retention (not margin erosion)
- Communication plan for program launch
Involuntary Churn Prevention
- Dunning email sequence (3-5 emails over 14 days)
- Smart retry logic for failed payments
- Card update reminder before expiration
Phase 4 — Output
- Deliver the retention playbook and engagement scoring spec.
- Run all email sequences through quality gates:
python E:\Dev2\kai-cmo-harness-work\scripts\quality_gates\banned_word_check.py <file>python E:\Dev2\kai-cmo-harness-work\scripts\quality_gates\four_us_score.py <file>
- Include a 90-day implementation roadmap and monthly metrics to track (churn rate, cohort retention, health score distribution, NPS trend, expansion vs. contraction revenue).
Constraints
- No banned Tier 1 words in any customer-facing copy.
- Win-back emails must comply with CAN-SPAM (reference:
E:\Dev2\kai-cmo-harness-work\harness\references\cold-email-rules.md).
- Loyalty program rewards must not erode margins below profitability.
- Discount offers in rescue plays capped at 20% unless user approves higher.
- All email sequences target 10+/16 on Four U's scoring.
- Max 2 auto-retry cycles on quality gate failures for email content.
1---2name: kai-retention-23description: Customer retention system — churn analysis, retention tactics, loyalty programs, and engagement scoring. Use when "retention", "reduce churn", "keep customers", "loyalty program", "customer retention", "churn prevention", "churn analysis", "engagement scoring", "win-back", "customer lifetime value", or any request to analyze, prevent, or reduce customer churn.4---5
6# kai-retention — Customer Retention System
7
8Design a complete retention system: churn diagnostics, retention tactics, engagement scoring, loyalty mechanics, and win-back campaigns.
9
10## Phase 0: Load Product Context
11
12Check if `MARKETING.md` exists in the **project root** (same directory as CLAUDE.md, README.md, package.json).
13
14**If it exists:** Read it — skip product discovery questions. It has the product name, ICP, value prop, monetization, brand voice, current channels, and competitive landscape.
15
16**If it does NOT exist:** Auto-explore the codebase to create it in the **project root** (next to CLAUDE.md). Do NOT ask the user what the product is. Read CLAUDE.md, README.md, PROJECT.md, package.json, landing pages, and any project files. Search for email/ad/analytics config. Then create `MARKETING.md` using the template from `/kai-email-system`. Present draft to user for confirmation.
17
18---
19
20## References
21
22Load these files as context before starting:
23
24- `E:\Dev2\kai-cmo-harness-work\knowledge\playbooks\customer-retention.md`
25- `E:\Dev2\kai-cmo-harness-work\knowledge\playbooks\growth-loops-applied.md`
26- `E:\Dev2\kai-cmo-harness-work\knowledge\channels\email-lifecycle.md`
27- `E:\Dev2\kai-cmo-harness-work\knowledge\personas\_persona-index.md`
28
29## Phase 1 — Discovery
30
311. Read from `MARKETING.md`. Only ask about things not covered there:
32 - Business model (SaaS, ecommerce, services, marketplace)
33 - Current churn rate (monthly/annual, if known)
34 - Customer count and average revenue per customer
35 - Current retention efforts (any emails, loyalty program, support)
36 - Known churn reasons (from exit surveys, support tickets, cancellation flow)
37 - Product usage data availability (do they track feature adoption?)
38 - Customer segments (free vs. paid, plan tiers, cohorts)
392. Identify the retention maturity level:
40 - **Level 0**: No retention effort beyond the product itself
41 - **Level 1**: Basic cancellation flow + occasional check-in emails
42 - **Level 2**: Lifecycle emails + usage tracking + support triggers
43 - **Level 3**: Predictive churn scoring + proactive intervention + loyalty program
44
45## Phase 2 — Analysis
46
47### Churn Diagnostics
481. Categorize churn types:
49 - **Voluntary**: Customer actively cancels (dissatisfaction, budget, switched)
50 - **Involuntary**: Payment failure, expired card, billing issue
51 - **Passive**: Stops using but doesn't cancel (ghost users)
522. Map the churn timeline: when do most customers leave?
53 - First 30 days (onboarding failure)
54 - 60-90 days (value not realized)
55 - At renewal (annual plan decision point)
56 - After price increase or feature change
573. Identify leading indicators of churn:
58 - Login frequency decline
59 - Feature usage drop
60 - Support ticket volume spike
61 - NPS/CSAT score decline
62 - Billing page visits
63
64### Engagement Scoring Model
65Define a health score (0-100) based on:
66
67| Signal | Weight | Scoring |
68|--------|--------|---------|
69| Login frequency (last 14 days) | 25% | Daily=100, Weekly=60, Monthly=20, None=0 |
70| Core feature usage | 25% | Used all=100, Used some=50, Used none=0 |
71| Support interactions | 15% | Positive=80, Neutral=50, Negative=20 |
72| Account expansion signals | 15% | Upgraded=100, Stable=50, Downgraded=10 |
73| Referral/advocacy | 10% | Referred=100, NPS promoter=60, Passive=30 |
74| Billing health | 10% | Current=100, Late=30, Failed=0 |
75
76**Risk tiers**: Green (70-100), Yellow (40-69), Red (0-39).
77
78## Phase 3 — Produce
79
80Build these deliverables:
81
82### Retention Playbook
83Intervention strategies by risk tier:
84
85**Red (0-39) — Immediate Rescue**
86- Trigger personal outreach within 24 hours
87- Offer concession (discount, extended trial, premium support)
88- Escalate to customer success manager
89- Deploy win-back email sequence
90
91**Yellow (40-69) — Proactive Nurture**
92- Send usage tips targeting unused features
93- Invite to office hours or webinar
94- Share relevant case study or success story
95- Request feedback (short survey, not NPS)
96
97**Green (70-100) — Expansion & Advocacy**
98- Request referral or testimonial
99- Offer early access to new features
100- Invite to advisory board or beta program
101- Cross-sell or upsell relevant add-ons
102
103### Win-Back Campaign
104For customers who have already churned:
105- 3-email sequence: Day 1, Day 7, Day 30
106- Each email addresses a different churn reason
107- Include a specific offer or product update
108- Run through quality gates before sending
109
110### Loyalty Program Design (if applicable)
111- Reward mechanics: points, tiers, milestones, or referral credits
112- Earning actions mapped to business goals
113- Redemption options that drive retention (not margin erosion)
114- Communication plan for program launch
115
116### Involuntary Churn Prevention
117- Dunning email sequence (3-5 emails over 14 days)
118- Smart retry logic for failed payments
119- Card update reminder before expiration
120
121## Phase 4 — Output
122
1231. Deliver the retention playbook and engagement scoring spec.
1242. Run all email sequences through quality gates:
125 - `python E:\Dev2\kai-cmo-harness-work\scripts\quality_gates\banned_word_check.py <file>`
126 - `python E:\Dev2\kai-cmo-harness-work\scripts\quality_gates\four_us_score.py <file>`
1273. Include a 90-day implementation roadmap and monthly metrics to track (churn rate, cohort retention, health score distribution, NPS trend, expansion vs. contraction revenue).
128
129## Constraints
130
131- No banned Tier 1 words in any customer-facing copy.
132- Win-back emails must comply with CAN-SPAM (reference: `E:\Dev2\kai-cmo-harness-work\harness\references\cold-email-rules.md`).
133- Loyalty program rewards must not erode margins below profitability.
134- Discount offers in rescue plays capped at 20% unless user approves higher.
135- All email sequences target 10+/16 on Four U's scoring.
136- Max 2 auto-retry cycles on quality gate failures for email content.
majiayu000