Best Practices & Troubleshooting
Best Practices
-
Provide Sufficient Historical Data: Set a Training Start Date covering at least 90–180 days of transaction data to ensure high-accuracy model predictions.
-
Filter Restricted SKUs: Set Exclusion Filters for out-of-stock items, internal store charges, or seasonal clearance goods that should not be automatically recommended.
-
Utilize Fallbacks: Always define a valid Fallback Lookback Days period to avoid empty recommendation slots for brand-new customers without prior purchase history.