Enter a new quote's real details below. This runs the actual trained GBM model (gbm_model.txt), SHAP, and the real suggest_premium.py rule layer — the same pipeline used throughout this project.
Annual: add each good carried, its merchandise group, and its % of shipment value (percentages for goods in the same group are summed to get that group's total %).
Single-trip (Avulsa): just select which goods are carried — no magnitude is needed. The model only ever uses each good's % share of the shipment, never a raw dollar amount, so we auto-split evenly across however many goods you add (same pattern as Transport Mode above).
Domestic (TN/RCTR-C): add each origin → destination Brazilian region leg and its % of shipment value.
RCTR-C is road-only in the real data (13,247 of 13,247 historical policies use 100% road, no other mode) — fixed automatically, not editable.
Real backend rule for Avulsa (Single-trip): just select which modes were used (yes/no) — no % split is asked of the user. We auto-split evenly across your selected modes behind the scenes so the model still receives a value.