Success story
Forecast Pharmaceutical Demand in a Changing Market
Use Rulex’s AutoML to capture seasonality and incorporate external demand signals, helping supply chain teams anticipate market needs with greater accuracy.
average error on monthly
forecasts of key products
forecasting horizon
best-fit model
selection
updated
with the latest
available data
A global pharmaceutical company needed to improve demand forecasting across several flagship products, where historical sales alone could not fully explain changes in market demand. Seasonal patterns, geographic differences, and external factors such as the spread of seasonal viruses could all influence what products were needed, where, and when.
The challenge was to incorporate these different demand signals into the forecasting process while maintaining the accuracy needed to balance product availability against the cost of carrying excess inventory.
Rulex enabled the company to combine historical sales with temporal, geographic, and external data, using machine learning to identify the patterns influencing demand and generate more accurate forecasts.
Building a forecast that evolves with demand
Bring the right demand signals together
The forecasting process combined sales data from different products with external variables such as temporal trends, geographic information, and the spread of seasonal viruses. Rulex Platform handled data import, integration, and cleansing within the same no-code environment, preparing the information required for modeling.
Find the best model for the data
With the data prepared, Rulex AutoML automated model selection rather than relying on a single predetermined forecasting approach. Different options could be tested and executed in parallel, with the best-fit model selected automatically according to the data being analyzed.
Auto-regressive techniques were also applied to capture periodic patterns, allowing the forecasting process to account for recurring variations in demand over time.
Keep forecasts aligned with changing demand
Forecasts were generated over a three-month horizon, both at aggregate and monthly level. Periodic re-runs and a sliding-window approach incorporated the latest available data into subsequent forecasts, helping maintain performance as demand patterns evolved and resulting in very low Mean Absolute Percentage Error (MAPE) scores.
The results then fed directly into downstream processes such as inventory replenishment and logistics, without requiring additional tools.
The impact
- Average error below 10% on monthly forecasts of key products
- Three-month visibility into expected product demand
- Forecasts kept current as new data and demand trends emerged
- Forecasting outputs integrated into existing inventory and logistics systems
- Domain experts retained visibility and control throughout the forecasting process
Accurate forecasting is one part of a broader supply-demand balancing process. Explore how forecasting, inventory management, collaborative planning, and automation work together to help organizations respond to changing demand.
Read the full article: The Supply Chain Balancing Act: Demand Planning and Forecasting in Action
Tech point
Take a closer look at the Rulex Platform capabilities behind this demand forecasting solution
01. Data orchestration for forecasting
Forecasting models depend on data that rarely comes from a single source. Rulex Platform can collect information from local and remote files, databases, enterprise systems such as SAP, cloud storage, and other repositories, then convert it into a unified tabular structure for aggregation, reconciliation, and transformation. This allows different types of historical, operational, and external data to enter the same analytical workflow without requiring a separate integration layer.
02. Guided AutoML
Rulex AutoML automates key stages of the machine learning workflow, including data preparation and pre-processing, feature selection, algorithm selection, and model generation. Multiple models can be evaluated rapidly, while more than 100 customizable variables and a broad selection of algorithms allow users to refine the process when greater control is required.
Unlike a black-box AutoML process, each task and its outputs remain visible throughout the flow. Users can inspect results, identify where adjustments may be needed, and compare model performance while retaining control over the analytical process.
03. Continuous forecasting
Rulex Platform supports the operationalization of forecasting models through scheduled and repeatable execution workflows. Data preparation, model execution, and prediction generation can be orchestrated within the same flow and rerun automatically as new data becomes available, reducing the need for manual intervention between forecasting cycles.
This allows forecasting processes to move beyond one-time model development and operate as repeatable production workflows, with updated inputs feeding subsequent model runs and outputs made available to downstream processes.
Why Rulex for demand forecasting
Demand forecasting is not only about choosing an accurate model. It requires reliable data, the flexibility to evaluate different modeling approaches, and a process that can keep pace as new information becomes available. Rulex brings these capabilities into a single environment, from data orchestration and AutoML to recurring model execution and downstream integration.
Explore more supply chain case studies powered by Rulex Platform
Optimize Inventory Replenishment Alongside APS
Flexibly automating planning decisions, including exceptions.
End-to-end Transport Optimization for Truck Logistics
Generating end-to-end, long-term transport plans in under one hour.

Dynamic Deployment Planning
Agilely managing disruptions across complex supply networks.
Frequently asked questions
How does Rulex improve demand forecasting accuracy?
Rulex allows organizations to combine historical information with relevant external variables rather than relying purely on data related to past sales. AutoML can then evaluate different modeling options and select the best fit for the data. In this pharmaceutical use case, the resulting monthly forecasts for key products achieved an average error below 10%.
How does Rulex AutoML support demand forecasting?
Rulex AutoML automates time-consuming stages of the machine learning process, including data preparation, preprocessing, feature selection, and model selection. Different models can be evaluated quickly, while users retain visibility into individual tasks and results and can adjust the process when needed.
Can Rulex forecasts adapt as new demand data becomes available?
Yes. Forecasting workflows can be rerun as new information becomes available rather than remaining tied to a static model output. In this project, three-month forecasts were periodically rerun with new data, while a sliding-window approach captured the latest trends and helped maintain forecasting performance over time.