Shopify Admin Churn Risk Scorer

Purpose

Predicts which customers are at risk of churning by analyzing their purchase patterns against their historical buying frequency. Calculates an expected next-purchase date for each repeat customer, then scores churn risk based on how overdue they are. Read-only — no mutations.

Prerequisites

• Authenticated Shopify CLI session: shopify store auth --store <domain> --scopes read_orders,read_customers
• API scopes: read_orders, read_customers

Parameters

Safety

ℹ️ Read-only skill — no mutations are executed. Safe to run at any time.

Churn Risk Scoring Model

For each customer with min_orders or more purchases:

1. Average Purchase Interval (API) = total days between first and last order / (order_count - 1)
2. Days Since Last Order (DSLO) = today - last_order_date
3. Overdue Ratio = DSLO / API
4. Churn Risk Score (0-100):
   • Overdue ratio ≤ 1.0 → Score 0-20 (Active)
   • Overdue ratio 1.0–1.5 → Score 20-50 (Cooling)
   • Overdue ratio 1.5–2.5 → Score 50-80 (At Risk)
   • Overdue ratio > 2.5 → Score 80-100 (Likely Churned)
5. Customer Lifetime Value (CLV) = total spend / customer age in years × expected remaining years

Workflow Steps

1. OPERATION: orders — query Inputs: query: "created_at:>='<NOW - days_back days>'", first: 250, select createdAt, totalPriceSet, customer { id, email, firstName, lastName }, pagination cursor Expected output: All orders with customer association

2. Group orders by customer, calculate per customer:

   • Order dates (sorted chronologically)
   • Average purchase interval
   • Days since last order
   • Total spend
   • Order count

3. OPERATION: customers — query (enrichment) Inputs: Customer IDs for at-risk and likely-churned segments Expected output: Contact details, tags, total spend

4. Calculate churn risk score and classify into segments

5. Estimate revenue at risk = sum of (annual_spend × churn_probability) for at-risk customers

GraphQL Operations

graphqlCopy code
# orders:query — validated against api_version 2025-01
query OrdersForChurnAnalysis($query: String!, $after: String) {
  orders(first: 250, after: $after, query: $query) {
    edges {
      node {
        createdAt
        totalPriceSet { shopMoney { amount currencyCode } }
        customer {
          id
          email
          firstName
          lastName
          numberOfOrders
        }
      }
    }
    pageInfo { hasNextPage endCursor }
  }
}

graphqlCopy code
# customers:query — validated against api_version 2025-01
query AtRiskCustomers($ids: [ID!]!) {
  nodes(ids: $ids) {
    ... on Customer {
      id
      email
      firstName
      lastName
      totalSpentV2 { amount currencyCode }
      numberOfOrders
      tags
      createdAt
    }
  }
}

Session Tracking

Claude MUST emit the following output at each stage. This is mandatory.

On start, emit:

Copy code
╔══════════════════════════════════════════════╗
║  SKILL: Churn Risk Scorer                    ║
║  Store: <store domain>                       ║
║  Started: <YYYY-MM-DD HH:MM UTC>             ║
╚══════════════════════════════════════════════╝

After each step, emit:

Copy code
[N/TOTAL] <QUERY|MUTATION>  <OperationName>
          → Params: <brief summary of key inputs>
          → Result: <count or outcome>

On completion, emit:

For format: human (default):

Copy code
══════════════════════════════════════════════
CHURN RISK REPORT  (<days_back> days analyzed)
  Repeat customers scored:  <n>
  ─────────────────────────────
  Active (score 0-20):      <n> (<pct>%)
  Cooling (score 20-50):    <n> (<pct>%)
  At Risk (score 50-80):    <n> (<pct>%)   ⚠️
  Likely Churned (80-100):  <n> (<pct>%)   🔴

  Revenue at risk:         $<amount>/year

  Top at-risk by value:
    <name> (<email>)  Score: <n>  Last order: <date>  Lifetime: $<n>

  Output: churn_risk_<date>.csv
══════════════════════════════════════════════

Output Format

CSV file churn_risk_<YYYY-MM-DD>.csv with columns: customer_id, email, first_name, last_name, order_count, total_spent, avg_purchase_interval_days, days_since_last_order, overdue_ratio, churn_risk_score, risk_segment, expected_annual_value

Error Handling

Best Practices

• Pair with customer-win-back skill to take action on At-Risk and Likely Churned segments.
• Use with rfm-customer-segmentation for a more holistic view of customer health.
• High-value churning customers (top 20% by spend) should get personalized outreach.
• Export At-Risk segment to email marketing platform for automated win-back sequences.
• Adjust risk_threshold based on your product type: consumables (1.3), fashion (1.5), furniture (2.0).