Customer Analytics

Churn Risk Scoring

Identify which customers are about to churn by analyzing behavioral signals, then get a prioritized intervention list.

What This Skill Does

This skill analyzes customer behavior signals to identify who is at risk of churning before they actually leave. It evaluates purchase recency vs. typical purchase cycles, frequency trends, engagement changes, and support interactions to score each customer as High, Medium, or Low risk.

The output is a prioritized list of at-risk customers with specific intervention recommendations — whether to offer a discount, send personalized outreach, or recommend a product.

What You Need

Customer behavior data including: last purchase date, total number of purchases, typical purchase interval, recent email open rates, support tickets (if available).

Prompt Template

Copy this prompt, replace the [BRACKETED] placeholders with your data, and paste into Claude.

Analyze these customer behavior signals to identify churn risk. For each customer, evaluate: - Days since last purchase vs. their typical purchase interval - Trend in order frequency (increasing, stable, or declining) - Recent support tickets or complaints (if provided) - Email engagement trend (if provided) - Average order value trend Score each customer's churn risk as High / Medium / Low with a brief rationale. For the top 20 highest-risk customers, recommend a specific intervention: - "Discount offer" — for price-sensitive customers showing purchase decline - "Personal outreach" — for high-LTV customers with recent complaints - "Product recommendation" — for customers whose usual repurchase cycle is overdue - "Win-back sequence" — for customers likely already churned Include timing: when to reach out for maximum recovery probability. Here is my customer data: [PASTE YOUR DATA]

Example Output

Here is what Claude produces when you use this skill with sample data.

**Churn Risk Summary:** | Risk Level | Count | % of Base | Annual Revenue at Risk | |------------|-------|-----------|----------------------| | High Risk | 47 | 5% | $38,400 | | Medium Risk | 123 | 12% | $22,100 | | Low Risk | 830 | 83% | — | **Key Insight:** 3 of your top 5 at-risk customers have unresolved support tickets. Resolve these BEFORE sending discount offers — fixing service issues yields 2-3x better retention than discounts for complaint-driven churn. **Top 10 Priority Interventions:** | # | Customer | LTV | Overdue | Signal | Action | When | Channel | |---|----------|-----|---------|--------|--------|------|---------| | 1 | C0234 | $1,240 | 45 days | Frequency dropped 60%, 1 ticket | Resolve ticket + personal outreach | Immediate | Phone + email | | 2 | C0891 | $890 | 30 days | 2 support tickets, email opens 5% | Fix issues + "we fixed it" + 15% off | Within 48h | Support + email | | 3 | C0445 | $670 | 60 days | Email opens at 0%, AOV declining | SMS win-back: 20% off favorites | This week | SMS (email dead) | | 4 | C0312 | $580 | 52 days | Purchase interval 2.3x normal | "Running low?" replenishment reminder | This week | Email | | 5 | C0178 | $520 | 38 days | AOV dropped 40% last 2 orders | Product quiz → personalized bundle rec | Within 7d | Email | | 6 | C0667 | $490 | 70 days | No engagement, previously 4x/yr buyer | Final SMS: 25% off, one-click reorder | This week | SMS only | | 7 | C0823 | $440 | 42 days | Support ticket + declining opens | Resolve ticket + curated recommendations | Within 48h | Email + SMS | | 8 | C0156 | $380 | 55 days | Browsed 3x without buying last month | Retarget with viewed-product ad + 10% off | Within 7d | Meta ads | | 9 | C0934 | $350 | 48 days | Email opens dropped 80% → 15% | Re-engagement: "New arrivals since you left" | Within 7d | Email | | 10 | C0401 | $310 | 65 days | Subscription cancelled last month | Personal email: "Why did you cancel?" survey | Within 7d | Email | **Projected Recovery:** At 25-30% win-back rate on top 10, expect to recover $1,400-$1,700 in annual revenue within 30 days.

Tips for Best Results

The most predictive churn signal is typically a decline in purchase frequency relative to that specific customer's pattern — not an absolute threshold.

Combine this with your RFM analysis — High Risk + High Historical Value customers should get VIP-level intervention.

Re-run weekly to catch at-risk customers early, before they fully disengage.

Skip the Prompt — Automate This

Finsi runs this analysis automatically on your live data. No prompting, no copy-pasting — just real-time insights and AI-powered recommendations.

See Retention Intelligence

Related Skills

Continue building your AI-powered e-commerce toolkit.

Get These Insights Automatically

Finsi connects to your Shopify, Stripe, and marketing tools to run these analyses in real time — no prompts needed. Get AI-powered recommendations delivered to your dashboard every day.