10 times faster automated demand forecasting for global pharma supply chain
Discover how we replaced 7 days of manual forecasting by 400 planners with a fully automated 1.5-hour ML pipeline and cut multi-million dollar losses.
Up to 5% increase in forecast accuracy
thanks to Machine Learning, neural networks, and structured data pipelines
Up to $50M in annual waste reduced
by replacing manual demand planning with automated, high-accuracy forecasting
10× faster supply forecasting process
with automated predictions for thousands of medical product and location pairs
Client results
We built a scalable pipeline by combining historical, contextual, and logistics data into one automated system — improving planning across supply chain, inventory, and operations.
This allowed the company to:
Ability to ship medical components and tests based on actual demand, for example during a viral outbreak — enabled by a self-learning forecasting system that updates itself monthly.
Replacement of a week-long manual supply forecasting process ran by 400 planners with a fully automated 1.5-hour workflow, allowing experts to focus on critical exceptions.
Higher forecast accuracy, consistently reaching 80%, significantly reducing medical components waste and unlocking up to $50 million dollar savings across global supply chain operations.
Our challenges
- Manual forecasting for 10,000+ products by 400+ planners
- Low forecast accuracy causing supply chain inefficiencies
- No automation or scalable infrastructure for demand prediction
- Disconnected data sources and lack of ML-ready datasets
- No monitoring of forecast quality or model performance
The client, a global pharmaceutical supplier, relied on manual forecasting to determine how many medical components and test kits to ship to hundreds of facilities worldwide. This process was time-consuming, inconsistent, and failed to meet the desired level of accuracy — resulting in inefficiencies and annual losses reaching millions of dollars.
End-to-end ML forecasting pipeline
We built a forecasting system that generated monthly supply predictions for over 6,000 products across thousands of global locations using statistical models and neural networks. The pipeline ingested historical and contextual data from SAP BW into Snowflake, prepared ML-ready datasets, trained and selected the best-performing models, and sent forecasts directly back to SAP
Scalability without increasing cost or complexity
The automated forecasting system covered 60% of the company’s global portfolio, without increasing team size or requiring duplicate efforts across local markets. As new components and facilities locations are added, the same pipeline can be extended without additional tooling or manual planning, enabling global rollout without proportional cost growth.
Adapting to change — every month
The forecasting system retrained models automatically each month using the latest available data, including recent demand trends, logistics constraints, and contextual signals. This made it possible to detect and respond faster to sudden changes in local conditions — such as virus outbreaks, regional supply chain disruptions, or seasonal surges — without manual intervention or rework.
Solution architecture
We developed a modern, secure, and scalable data strategy tailored to both current infrastructure and long-term business goals.
See more of our work
.png)
Looking for a deeper look into our Data Platform architecture?
Use data to guide your business decisions towards better results
From managing your customer acquisition and retention, to product optimisation - we can help!