Demand Forecasting of Consumer Goods for the Indian Subcontinent

Our sponsor company, a leading fast-moving consumer goods corporation in the Indian Subcontinent (ISC), faces ongoing challenges in forecasting the volatile demand in this region. Currently, company’s demand planning relies heavily on human judgment, resulting in persistent inaccuracies and biases. This capstone develops a standardized and quantitative demand forecast methodology through three explorations. First, we build a variety of time series forecasting models. Second, we include exogenous variables relevant for Indian demand, such as rainfall and Consumer Price Index (CPI). Third, we explore ways to mitigate the impact of data outliers during the COVID-19 pandemic. These enhancements reduce the average mean absolute percent error (MAPE) of demand forecast across all product categories to 8.0%. We consolidate our analysis and model selection algorithms into an application named The Demand Forecaster. This powerful tool not only enables the sponsor company to retrieve our existing forecast results for all product categories, but also allows them to perform forecasts in the future with updated demand data and model specifications. By adopting our demand planning methodology, the sponsor company is expected to increase margins with better inventory management and production planning.