Fix: Remove customer_id causing Cartesian product in monthly loan aggregations

[jjoyce0510]

🐛 Bug Fix: Data Duplication in Risk Analytics Dashboard

This PR fixes the critical data quality issue causing inflated loan values in the Risk Analytics Reporting Dashboard.

---

Problem Summary

PR #3 (merged Jan 27, 2026) introduced a Cartesian product bug that duplicated aggregated loan data:

• ❌ Dashboard showing way higher loan values than previous months
• ❌ Row count assertion failing continuously (10/10 recent checks)
• ❌ Expected 10-30 rows, getting hundreds instead

---

Root Cause

The bug was in this join logic added by PR #3:

left join loans
    on orig.loan_type_name = loans.loan_type_name  -- ⚠️ No unique constraint!

What happened:

• Each monthly aggregate (e.g., "Mortgages - January 2026")
• Got multiplied by EVERY customer with that loan type
• If 50 customers have mortgages → 1 row becomes 50 duplicate rows
• Dashboard sums these duplicates → loan amounts inflated 50×

Why customer_id doesn't belong here:

• This table aggregates data by month + loan type
• One month has multiple customers → adding customer_id breaks the aggregation grain
• You can't have one customer_id per aggregated row

---

The Fix

✅ Removed customer_id field
✅ Removed LEFT JOIN loans causing duplication
✅ Restored original aggregation logic (pre-PR #3)

---

What Changed

combined as (
    select
        coalesce(orig.month_start, pay.month_start) as month,
        orig.loan_type_name,
-       loans.customer_id,  ⬅️ REMOVED
        coalesce(orig.loans_originated, 0) as new_loans,
        coalesce(orig.total_amount_originated, 0) as amount_originated,
        ...
    from monthly_originations orig
    full outer join monthly_payments pay
        on orig.month_start = pay.month_start
-   left join loans                      ⬅️ REMOVED
-       on orig.loan_type_name = loans.loan_type_name  ⬅️ REMOVED
)

---

Impact

✅ Fixes: Risk Analytics Reporting Dashboard will show correct loan values
✅ Fixes: Row count assertion will pass (returns to expected 10-30 rows)
✅ Fixes: Accurate monthly loan aggregations restored

---

Testing Recommendations

After merge:

1. Re-run dbt model: dbt run --select agg_monthly_loans
2. Verify row count is back to ~10-30 rows
3. Check dashboard values match pre-Jan 27 trends
4. Confirm row count assertion passes

---

Related

• Introduced by: PR Add customer_id to monthly loan aggregations #3 (Jan 27, 2026)
• Reported by: john.joyce@acryldata.io
• Affected dashboard: Risk Analytics Reporting (Tableau)