70% better accuracy for demand forecasting
Developed a new demand forecasting solution from scratch, that reduced the client’s time to market, optimized inventory management, and improved sales consultant experience by reducing stockouts
The client is a major cosmetics company in Latin America, operating under the direct sales model via magazines and catalogs in over 14 countries. They wanted to create a demand forecasting solution for their digital consultants that catered to the requirements of both the campaign management and supply chain teams. The client was looking to:
- Increase sales
- Reduce stock-outs and inventory
Sigmoid provided end-to-end solutions to the client from data preparation to creating a model using a stacked approach combining XG Boost and random forest. For the data preparation phase, Sigmoid used data from physical catalogs and digital offers to create a combined raw dataset and capture behavior trends of potential customers. Further, the solution was deployed by building ETL pipelines, automated training components, prediction components, and testing before migration to the client production environment. The model was further automated and scaled using Spark-based architecture.
Sigmoid automated the overall campaign planning solution for digital channels while improving the existing demand forecast solution by 2.7x prediction accuracy. It led to over $30 million savings in inventory handling costs and improving the experience of digital consultants by reducing stockouts.