For one of the largest package service providers in North America, running more than 600 facilities, 40+ hubs, and over 100,000 ground vehicles, every delivery is a small promise: get this package where it needs to go, safely and on time. Keeping that promise at scale meant constantly working to improve delivery excellence, customer experience, and driver safety through smarter logistics. But there was a gap sitting right at the heart of the operation. The company couldn’t accurately predict how many ground packages it would need to move by category, which meant forecasting the volume of different shipment types was more guesswork than science. Without that clarity, resource allocation and route planning were harder than they needed to be, and the ripple effects touched everything from delivery times to driver workloads. Compounding the problem, the teams responsible for building and managing the machine learning models meant to solve this weren’t working from a shared, reliable process, and a lack of collaboration between them was slowing down the very models that were supposed to bring the business more certainty. The company needed a partner who could bring both the data engineering depth and the collaborative discipline to turn forecasting from a weak point into a genuine advantage.
Ascendion played the role of a strategic partner, bringing engineering capacity and deep data engineering expertise in DataOps and MLOps to help the client leverage modern data strategies and drive technology expertise across the business.
Ascendion implemented Databricks MLOps with an MLflow solution using Ascendion AVA’s ML Optimization studio, using forecast models such as ARIMA, Prophet, Exponential Smoothing, and XGBoost within a single, unified framework. MLflow’s tracking capabilities were deployed to enable easy comparison and evaluation of the forecast models, with monitoring and logging mechanisms in place to track the performance of every deployed model. Automation and continuous integration helped refine the models, optimize resource allocation, identify bottlenecks, and reduce delivery delays, resulting in improved customer satisfaction. Forecast models are continuously refined based on user feedback and changing factors like weather and traffic.
Tech Stack:
Enabled the client to deliver 40% faster across the 3 billion packages it handles per year, with the data engineering solution improving forecasting that optimized resource allocation and route planning.
Reduced deployment time by 65%.
Increased volume forecast accuracy by 40%.
Delivered an 80% insights gain for better resource allocation and route planning.
Achieved a 40% improvement in delivery time, while enabling proactive notifications.
Enabled real-time response for proactive decision-making.
