Freight Rate Predictor
What should this load pay?
Pick a lane, equipment and date. The model prices it from 48,000 historical truckload shipments.
Your quote will appear here.
Price many loads at once
Upload a CSV with columns pickup, delivery, equipment, date and optionally
weight, distance, market_index. You get the same file back with a predicted_rate column.
How the model works
A linear regression on log(rate) carries distance, weight, market and the month-on-month price drift. A LightGBM model then corrects what the line misses, mostly lane geography. Cities are encoded by coordinates, so the model also prices lanes it has never seen.
Tested the honest way: trained on January–August 2025, scored on September–October 2025, which it never saw.
| Model | Avg. error | Avg. % error | R² |
|---|
Predicting the average rate for every load would be off by $1,183 on average; the final model is about 11× closer.