Background
A leading global sportswear and athletic footwear brand needed to modernize its demand forecasting to manage a complex supply chain, diverse product categories, and frequent product launches. The company’s existing manual forecasting methods offered limited accuracy and lacked the agility required for fast-paced market dynamics.
Solution
Affine developed a fully automated, AI-powered demand forecasting engine tailored to the client’s planning needs: Used LSTM-based deep learning models at the customer-category level, incorporating POS and macroeconomic signals Applied ensemble ML models (e.g., Random Forest, XGBoost) for style-level forecasting, including for new product launches Implemented a champion model selection framework based on sales volume and product lifecycle stage Reduced forecasting cycle time from 7 days to 3 hours through full pipeline automation Delivered outputs via a self-serve web dashboard, enabling planners and merchandisers to access forecasts and simulate demand scenarios in real time
Impact
Increased forecast accuracy from 60% to 90% 90% reduction in forecast generation time (from 7 days to 3 hours) Achieved full SKU coverage, including low-history and new product items Enabled faster, data-driven inventory and launch decisions Improved product availability and customer satisfaction