ic_003 · Industry: E-commerce — repeat purchase (30-day)
Status: ✓ Passing · Last run: 2026-06-07T07:40:43 · Pipeline: v2
· History: 1/2 runs passed (50%)
What was tested
E-commerce customer retention: predict 30-day repeat-purchase per customer. Five tables: customers + product_catalog (dim) + sessions + transactions + labels. Planted signal: recent purchase recency + session frequency + product variety drive repeat purchase.
What we planted in the data
- 300 entities
- Signal strength: 0.70 (sigmoid slope multiplier)
What the system did
- Training rows built: 1,200
- Features used in model: 29
- Model selected:
LogisticRegression - CV AUC: 0.769
Top features by importance:
session_count_180d(0.198)txn_amount_sum_365d(0.166)txn_count_roc_7_30(0.105)region_US(0.104)txn_amount_sum_7d(0.095)
What we checked — all assertions passed
| Status | Assertion | Detail |
|---|---|---|
| ✓ | no_pipeline_errors |
clean run |
| ✓ | feature_survival_rate_above |
29/44 features survived (66% ≥ 40%) |
| ✓ | source_tables_contributing_above |
4 distinct source-table prefixes in surviving features (['pages', 'region', 'session', 'txn']) |
| ✓ | auc_in_band |
AUC = 0.769 inside band [0.55, 0.95] |
How this could have gone wrong
(no assertion descriptions provided)
Why this case matters
E-commerce repeat-purchase prediction is one of the most common "marketing analytics" asks. The shape (customers + sessions + transactions + product dim) covers most retail data layouts.
Reproducing
# from auto_insight_api/
python -m validation.v2 run ic_003 --pipeline v2 -v
- Case config:
validation/v2/cases/ic_003_ecommerce_repeat_purchase.yaml - Data shape:
industry_ecommerce_repeat_purchase - Analytics type:
ml_binary