The coronavirus pandemic has caused unprecedented demand and supply shocks and lasting changes in consumer attitudes and preferences. Historical demand patterns, traditional time-series and forecasting methods are ineffective in the current economic state with vast disruptions and unpredictable inflection points. We present forecasting models using ML and NLP to provide actionable decision support
The coronavirus pandemic has caused unprecedented demand and supply shocks and lasting changes in consumer attitudes and preferences. In the post-COVID "next normal" retailers and brands must anticipate volatility and build demand sensing capabilities to understand myriad signals, assess their impact on demand, supply, and operations, and take action in "real time" across the value chain and extended ecosystem.
Historical demand patterns, traditional time-series and forecasting methods are ineffective in the current economic state with vast disruptions and unpredictable inflection points. There is a need for more accurate forecasts at the product, channel and locality level as well as real-time insights to improve and accelerate demand sensing decision making.
We present a demand sensing solution that tracks implications and consequences of disruptions, using machine learning and natural language processing to provide a 360 degree view of demand shifts and early warnings to supply chain risk. We use feature engineering and analytical modeling to derive signals of pandemic resurgence, economic volatility, and consumer safety sentiment. These features are embedded into forecasting models to provide actionable decision support. We have deployed our solution at numerous retailers to predict demand at the most granular geographic levels for their most volatile product categories.