Improved Forecast
Accuracy by 35%
Reduced Man-hours by
30%
Enabled What-If Analysis
Our client is a major producer and distributor for various agriculture and food products like oil seeds, dairy feed, cooking oil, etc. While the organization operates globally, the meats business unit provides a range of meat in a B2B business model to various large food chains across North and South America.
The client was facing challenges in its Demand Planning function, majorly due to siloed planning, gaps in processes like effective segregation of promotions and inappropriate forecast models, as well as Master Data issues.
The aim was to reinvent the entire Demand Planning process through a modern planning platform that would harmonize not only the process but also improve the effectiveness and accuracy of the Demand Planning outputs.
Thoucentric team utilized its knowledge of Demand Planning best practices as well as technical expertise to establish a digitally powered demand planning process by studying the business process, analysing the gaps and challenges, and driving a successful deployment.
The team enabled the use of standard functionalities like New Product Launch, Product Segmentation and Seasonal Forecasting as well as out-of-the-box solutions for pattern recognition and forecast model assignment, short-term forecasting using Open Sales Order, Shipments and POS Data. Forecasts were further refined by incorporating promotions and macroeconomic trends. A standardized set of Forecast Models, including ML-based models like ARIMAX were created to be mapped to the appropriate product categories.
While the appropriate training was given to planners at the onset, quick reference guides were also provided for future use.
The demand planning suite consisting of numerous forecasting models, assigned based on patterns identified in historical data, helped improve the efficiency while providing flexibility. Consideration of product categorization, seasonality and new product launch scenarios further added value to the forecasting process. Promotions and trends were utilized to refine the forecast and near-term forecast was generated using demand signals like Point of Sales and shipment.
ARIMAX Model – This is an advanced multi-linear regression model that considers the stationarity of data and the effects of external variables. As part of the implementation, the ARIMAX model was deployed, which considers the effects of macroeconomic factors like GDP, Industry Price of meat, Market Share, etc., which resulted in a further improvement in forecast accuracy.