shopify-admin-churn-risk-scorer

What this skill does

## 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

| Parameter | Type | Required | Default | Description |
|-----------|------|----------|---------|-------------|
| store | string | yes | — | Store domain |
| days_back | integer | no | 365 | Historical window for purchase pattern analysis |
| min_orders | integer | no | 2 | Minimum orders to calculate purchase interval (need 2+ for frequency) |
| risk_threshold | float | no | 1.5 | Multiplier of avg purchase interval before flagging as at-risk |
| format | string | no | human | Output format: `human` or `json` |

## 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

```graphql
# 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 }
  }
}
```

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

**After each step**, emit:
```
[N/TOTAL] <QUERY|MUTATION>  <OperationName>
          → Params: <brief summary of key inputs>
          → Result: <count or outcome>
```

**On completion**, emit:

For `format: human` (default):
```
══════════════════════════════════════════════
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
| Error | Cause | Recovery |
|-------|-------|----------|
| `THROTTLED` | API rate limit exceeded | Wait 2 seconds, retry up to 3 times |
| Single-purchase customers | Can't calculate interval | Exclude from scoring (need 2+ orders) |
| Guest orders | No customer linkage | Skip — cannot build customer profile |

## 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).