\r \r

Customer Churn Prediction Analyst\r

\r

Overview\r

\r The Customer Churn Prediction Analyst is a production-grade intelligence tool that identifies at-risk customers before they leave. By analyzing multi-dimensional behavioral signals—purchase frequency trends, support ticket sentiment, feature adoption rates, engagement decay, and payment friction—this skill surfaces customers most likely to churn within 30/60/90 days.\r \r Beyond prediction, it generates actionable intervention playbooks: personalized discount strategies, feature education campaigns, re-engagement email templates, and VIP outreach scripts. The skill integrates with Stripe (payment history, subscription metrics), Shopify (order patterns, product affinity), SaaS platforms (API usage logs, login frequency), and Slack (automated alerts for high-risk segments).\r \r Why it matters: Research shows that acquiring a new customer costs 5-25x more than retaining an existing one. A 5% improvement in retention can increase profitability by 25-95%. This skill automates the intelligence layer that turns data into revenue protection.\r \r ---\r \r

Quick Start\r

\r Try these prompts immediately:\r \r

Example 1: Analyze Stripe Subscription Churn Risk\r

Analyze my Stripe customer base for churn risk. \r
I have 1,200 active subscriptions ranging from $29-$299/month.\r
Look at: payment failures in the last 90 days, \r
declining MRR trends, and customers who haven't logged in for 30+ days.\r
Generate a risk-ranked list of my top 50 at-risk accounts \r
with specific intervention recommendations for each.\r
```\r
\r
### Example 2: Shopify E-commerce Customer Retention\r
```\r
I run a Shopify store with 8,500 customers. \r
Identify customers at risk of not returning.\r
Analyze: purchase frequency decline, \r
average order value trends, cart abandonment patterns, \r
and email engagement (bounces/unsubscribes).\r
Create win-back campaign templates for three risk tiers: \r
High (80%+ churn probability), Medium (50-79%), Low (25-49%).\r
Include personalized discount offers and subject lines.\r
```\r
\r
### Example 3: SaaS Feature Adoption & Engagement Churn\r
```\r
Analyze our SaaS platform for churn signals.\r
Our customers are: 120 paid accounts, \r
avg contract value $5,000/month.\r
Track: API call volume (declining usage = risk), \r
feature adoption (low-feature users churn 3x faster), \r
support ticket sentiment (negative = escalation risk), \r
and last login recency.\r
Flag accounts with \x3C10 API calls/week or \r
no logins in 14+ days as critical intervention targets.\r
Generate retention playbooks for each.\r
```\r
\r
---\r
\r
## Capabilities\r
\r
### 1. **Multi-Source Behavioral Analysis**\r
Aggregates signals from multiple platforms into a unified churn risk model:\r
\r
- **Stripe Integration:** Payment decline frequency, subscription downgrades, MRR trajectory, failed payment recovery attempts, dunning email effectiveness\r
- **Shopify Integration:** Purchase frequency (RFM: Recency, Frequency, Monetary), product category affinity, cart abandonment rate, average order value trends, customer lifetime value (CLV) projections\r
- **SaaS/API Platforms:** Daily active users (DAU), feature adoption rates, API call volume patterns, session duration trends, support ticket volume/sentiment, last-activity timestamps\r
- **Email/CRM Data:** Open rates, click-through rates, unsubscribe trends, email bounce rates, campaign engagement decay\r
- **Support Systems:** Ticket volume, resolution time, sentiment analysis (negative sentiment = 4x higher churn risk), escalation frequency\r
\r
### 2. **Predictive Risk Scoring**\r
Generates 30/60/90-day churn probability scores using:\r
\r
- **Recency Decay:** How long since last transaction/login (exponential weighting)\r
- **Frequency Trends:** Purchase/usage slope analysis (declining = risk signal)\r
- **Monetary Value:** Revenue-at-risk calculations; high-value customers flagged separately\r
- **Engagement Velocity:** Rate of engagement decline vs. historical baseline\r
- **Cohort Benchmarking:** Compare customer behavior to cohort norms (e.g., customers acquired in same month)\r
- **Seasonal Adjustment:** Account for industry seasonality (e.g., retail Q4 spikes)\r
\r
**Output:** Risk tiers (Critical, High, Medium, Low) with confidence intervals.\r
\r
### 3. **Personalized Intervention Recommendations**\r
Generates tailored win-back strategies:\r
\r
- **Segment-Specific Offers:** High-value customers get VIP treatment (white-glove support, exclusive features); price-sensitive get discounts; feature-poor get education\r
- **Email Campaign Templates:** Pre-written re-engagement sequences with A/B test variants, personalized product recommendations, and dynamic subject lines\r
- **Feature Education Playbooks:** For SaaS: identify underutilized features that correlate with churn; generate feature demo videos, webinar invites, or one-on-one training offers\r
- **Support Escalation Triggers:** Route customers with 3+ negative support interactions to dedicated success managers\r
- **Win-Back Incentive Suggestions:** Recommend discount depth (5%, 10%, 20%) based on customer LTV, willingness-to-pay analysis, and competitive benchmarking\r
\r
### 4. **Retention Campaign Orchestration**\r
Generates ready-to-deploy campaigns:\r
\r
- **Multi-Channel Sequences:** Email → SMS → In-App Push → Slack notification → Phone outreach (for high-value accounts)\r
- **Timing Optimization:** Send interventions at peak engagement windows (e.g., Tuesday 10am for B2B SaaS)\r
- **Dynamic Content:** Personalized product recommendations, usage statistics, and social proof ("3 customers like you upgraded to Pro this month")\r
- **A/B Test Frameworks:** Generate variant subject lines, offer amounts, and CTA copy for testing\r
\r
### 5. **Win-Back Success Tracking**\r
Monitors intervention effectiveness:\r
\r
- **Conversion Metrics:** % of at-risk customers who re-engage, upgrade, or extend contracts post-intervention\r
- **ROI Calculation:** Cost per intervention vs. revenue recovered; payback period\r
- **Cohort Analysis:** Which intervention types work best for which customer segments?\r
- **Feedback Loop:** Continuous model refinement based on what interventions actually prevent churn\r
\r
---\r
\r
## Configuration\r
\r
### Environment Variables (Required)\r
```bash\r
# Stripe integration\r
export STRIPE_API_KEY="sk_live_..."\r
\r
# Shopify integration\r
export SHOPIFY_API_TOKEN="shppa_..."\r
export SHOPIFY_STORE_NAME="your-store.myshopify.com"\r
\r
# SaaS/custom platform\r
export SAAS_API_KEY="your_saas_api_key"\r
export SAAS_API_ENDPOINT="https://api.yourplatform.com/v1"\r
\r
# OpenAI (for recommendation generation)\r
export OPENAI_API_KEY="sk-..."\r
\r
# Slack notifications (optional)\r
export SLACK_WEBHOOK_URL="https://hooks.slack.com/services/..."\r
\r
# Database (for tracking historical interventions)\r
export DATABASE_URL="postgresql://user:pass@localhost/churn_db"\r
```\r
\r
### Setup Instructions\r
\r
1. **Authenticate with data sources:**\r
   ```bash\r
   # Stripe: Generate API key from Dashboard > Developers > API Keys\r
   # Shopify: Admin > Apps and Integrations > Develop Apps > Create API credentials\r
   # SaaS: Use your platform's API documentation\r
   ```\r
\r
2. **Initialize the analysis:**\r
   ```bash\r
   # First run: full historical analysis (may take 5-10 minutes for large datasets)\r
   openclaw run customer-churn-prediction-analyst \\r
     --mode=full-analysis \\r
     --lookback-days=180 \\r
     --data-sources=stripe,shopify,saas\r
   ```\r
\r
3. **Set up recurring analysis:**\r
   ```bash\r
   # Schedule weekly churn analysis\r
   openclaw schedule customer-churn-prediction-analyst \\r
     --frequency=weekly \\r
     --day=monday \\r
     --time=08:00 \\r
     --notify-slack=true\r
   ```\r
\r
### Configuration Options\r
- `risk-threshold`: Churn probability threshold (default: 0.5 = 50%)\r
- `lookback-days`: Historical analysis window (default: 180 days)\r
- `prediction-horizon`: Predict churn within X days (default: 30, 60, 90)\r
- `high-value-threshold`: Revenue amount that triggers VIP intervention (default: $5,000 MRR)\r
- `intervention-budget`: Maximum discount/incentive per customer (default: 15% of CLV)\r
\r
---\r
\r
## Example Outputs\r
\r
### Output 1: Churn Risk Report (JSON)\r
```json\r
{\r
  "analysis_date": "2025-01-15T10:30:00Z",\r
  "total_customers_analyzed": 1247,\r
  "churn_risk_distribution": {\r
    "critical": 23,\r
    "high": 87,\r
    "medium": 156,\r
    "low": 981\r
  },\r
  "at_risk_accounts": [\r
    {\r
      "customer_id": "cust_8x9y2z",\r
      "name": "Acme Corp",\r
      "mrr": 12500,\r
      "churn_probability_30d": 0.89,\r
      "churn_probability_60d": 0.76,\r
      "primary_risk_signals": [\r
        "API usage declined 65% in last 30 days",\r
        "Payment failed 2x (recovered 1x)",\r
        "Support ticket sentiment: negative (3 tickets)",\r
        "No login in 18 days"\r
      ],\r
      "recommended_intervention": {\r
        "type": "VIP_SAVE",\r
        "tactics": [\r
          "Schedule executive business review call",\r
          "Offer 20% discount + feature unlock for 3 months",\r
          "Assign dedicated success manager"\r
        ],\r
        "estimated_recovery_probability": 0.72,\r
        "estimated_clv_at_risk": 150000\r
      },\r
      "suggested_email_subject": "We miss you, Acme—here's what's new in Q1"\r
    }\r
  ],\r
  "revenue_at_risk": 487500,\r
  "recommended_intervention_budget": 73125,\r
  "estimated_roi": 5.7\r
}\r
```\r
\r
### Output 2: Intervention Campaign Template\r
```markdown\r
## Re-Engagement Campaign: "Win Back Acme Corp"\r
\r
**Target Segment:** High-value SaaS customers with 60%+ churn risk\r
**Timing:** Send Monday 9am PT\r
**Duration:** 3-week sequence\r
\r
### Email 1: "We noticed you've been quiet"\r
Subject: Acme, we want to help—here's what's new [A/B variant: "Your exclusive preview inside"]\r
\r
Hi [FirstName],\r
\r
We noticed your team's API usage has dropped. That's usually a sign we haven't delivered enough value—and that's on us.\r
\r
**Here's what we've shipped since you last logged in:**\r
- Real-time collaboration (your #1 feature request)\r
- 40% faster query performance\r
- New integrations: Salesforce, HubSpot, Slack\r
\r
**Offer:** Upgrade to Pro free for 90 days + 1:1 onboarding session ($0 cost to you).\r
\r
[Claim Offer Button]\r
\r
Questions? Reply to this email or book time with Sarah, your success manager: [Calendly Link]\r
\r
---\r
\r
### Email 2: "Social Proof" (Day 5)\r
Subject: 3 customers like you switched to Pro this month—here's why\r
\r
[Testimonials, case study, usage stats]\r
\r
---\r
\r
### Email 3: "Final Offer" (Day 14)\r
Subject: Last chance: 25% off Pro + dedicated support [Expires Friday]\r
\r
[Time-limited offer, scarcity messaging]\r
```\r
\r
### Output 3: Win-Back Success Dashboard\r
```\r
Churn Prevention Dashboard (Last 30 Days)\r
\r
At-Risk Customers Identified:     156\r
Interventions Deployed:            143 (92%)\r
Re-Engaged (logged in post-email): 89 (62%)\r
Converted to Upgrade:              34 (24%)\r
Revenue Recovered:                 $47,300\r
Intervention Cost:                 $3,200\r
ROI:                               14.8x\r
\r
Top Performing Interventions:\r
1. VIP Phone Call (67% re-engagement rate)\r
2. Feature Education Webinar (58%)\r
3. Discount Offer (35%)\r
4. Email Sequence (28%)\r
```\r
\r
---\r
\r
## Tips & Best Practices\r
\r
### 1. **Segment Before Intervening**\r
Don't use one-size-fits-all offers. High-value customers respond better to white-glove service; price-sensitive segments respond to discounts. This skill auto-segments—use it.\r
\r
### 2. **Timing is Everything**\r
Send interventions during peak engagement windows. For B2B SaaS, that's usually Tuesday-Thursday, 9-11am. For e-commerce, Friday evening often works best. Test and adjust.\r
\r
### 3. **Feature Education Beats Discounts**\r
Customers who adopt 3+ core features have 10x lower churn. Before offering discounts, try feature education. It's cheaper and builds stronger retention.\r
\r
### 4. **Track the Tracking**\r
Set up UTM parameters and unique promo codes for each intervention so you can measure ROI. Example: `utm_source=churn_email&utm_medium=reengagement&utm_campaign=acme_save`\r
\r
### 5. **Weekly Monitoring Over Batch Processing**\r
Run churn analysis weekly, not monthly. Early intervention (when churn probability hits 40%) is 3x more effective than waiting until it hits 80%.\r
\r
### 6. **Validate Risk Signals Manually**\r
If the skill flags a high-value customer as high-risk, spot-check the data manually before sending a "we're losing you" message. False positives damage trust.\r
\r
### 7. **Personalize at Scale**\r
Use dynamic content blocks in emails. Instead of "Here's a discount," say "We noticed you use our Reports feature heavily—here's a 20% upgrade to Pro Reports."\r
\r
### 8. **Combine with Product Changes**\r
If the skill identifies that low feature adoption = churn, talk to product. Maybe the feature is hard to discover. Fix the product, not just the customer.\r
\r
---\r
\r
## Safety & Guardrails\r
\r
### What This Skill Will NOT Do\r
\r
1. **Discriminatory Targeting:** This skill will NOT use protected characteristics (age, race, gender, location) as churn risk factors. All recommendations are based on behavioral and transactional signals only.\r
\r
2. **Aggressive Dark Patterns:** This skill will NOT generate deceptive subject lines, fake urgency ("Only 2 left!"), or manipulative CTAs. All messaging is honest and customer-centric.\r
\r
3. **Unlimited Discounting:** Intervention budgets are capped per customer (default: 15% of CLV). The skill will NOT recommend discounts that would make the customer unprofitable.\r
\r
4. **Automatic Execution:** This skill generates recommendations; **you must approve all interventions before sending**. It will not auto-send emails or modify customer accounts without explicit approval.\r
\r
5. **Privacy Violations:** This skill respects GDPR, CCPA, and CAN-SPAM regulations. It will NOT:\r
   - Segment based on sensitive personal data\r
   - Send emails to unsubscribed users\r
   - Retain PII longer than necessary\r
   - Share customer data with third parties\r
\r
6. **Over-Reliance on Predictions:** Churn prediction models are probabilistic, not deterministic. A 89% churn probability doesn't mean the customer *will* churn. Use it as a signal, not gospel.\r
\r
### Limitations\r
\r
- **Data Quality Dependency:** Garbage in, garbage out. If your data is incomplete or inaccurate, predictions suffer. Ensure Stripe/Shopify/SaaS data is clean and current.\r
- **Cold Start Problem:** New customers (\x3C 30 days) don't have enough historical data for reliable churn prediction. The skill will flag these as "insufficient data."\r
- **Industry Variance:** Churn models are trained on general patterns. Your industry may have unique dynamics. Validate predictions against your domain knowledge.\r
- **External Factors:** Skill can't account for macroeconomic shocks, competitor actions, or