An automated freight cost prediction system was developed for a major home improvement retailer to handle half a million requests daily. Key challenges included data quality, digital infrastructure, variable shipping price factors, and performance requirements. A machine learning model using XGBoost was chosen for its accuracy and robustness. The system improved shipping cost predictions, increasing customer conversion rates and reducing underpricing risks. Lessons learned emphasized the importance of data quality and simpler solutions over complex models.
Table of contents
Project Scope and Technological ImplementationKey Challenges AddressedImplementation ApproachResults and Performance MetricsKey Lessons LearnedConclusion25 Impressions