SoyaFlow

SoyaFlow is a unified distribution and logistics platform built for Soya Excel, spanning operations across Canada, the United States, and Spain. It combines machine learning, GIS, and a modern web interface to coordinate soybean meal supply from inventory to farmer delivery.

Instead of installing expensive IoT bin sensors at every customer site, SoyaFlow uses an XGBoost model trained on 62 historical and operational features to predict when each farmer will reorder. Managers get reorder alerts, drivers get optimized multi-stop routes, and leadership gets KPI dashboards that reveal fleet efficiency and demand trends.

Predictive Reorder Modeling

A gradient-boosted tree model (XGBoost) learns from 62 engineered features, including historical order cadence, seasonality, geography, product mix, and livestock-cycle signals. The model reaches roughly 95% accuracy on reorder prediction, giving the scheduling team a week of forward visibility without physical sensors.

Route Optimization

Daily delivery routes are built by clustering farmer stops with DBSCAN and KMeans, then routed through Google Maps for drive-time and distance estimates. The system balances truck capacity, delivery windows, and clustering geometry to minimize empty miles.

Operational Workflow

The platform runs a weekly cycle (Tuesday planning, Friday finalization) with role-scoped views for managers, drivers, and client-facing staff. Orders move through a pending-to-delivered workflow with batch and expedition tracking, while inventory and low-stock alerts keep production synced with demand.

Sustainability

Scope 3 emissions tracking across the distribution fleet turns logistics data into a sustainability lens, supporting procurement and reporting requirements alongside operational efficiency.