The client is a global healthcare and pharmaceuticals company and wanted to create a demand forecasting solution that works for different categories of products. Their existing system lacked a single source of true forecast that could be consumed across different business teams. They needed to shorten the current forecasting cycle and make the process more efficient.
Sigmoid improved their existing forecast solution using ML algorithms to improve WAPE, BIAS, and coverage. The process involved removing the seasonality and trend from the products to get base sales. The algorithm catered to different patterns in the data like non-linearity, high volatility, extreme values, non-stationarity, sudden peaks, and troughs, etc. Sigmoid also developed new models using ML algorithms for new product launches using SKU-SKU similarity and pattern matching based on their sales values.
Created an MVP for 3 product categories with improved accuracy of demand forecasting models by 30 percentage points.