The Role of AI and Retail Market Knowledge in Predicting Demand
Retail Tech Insights | Monday, March 02, 2020
AI/ML will continue to improve forecasting capabilities, resulting in more robust decision-making processes that better satisfy customer needs.
FREMONT, CA: The Covid epidemic wreaked havoc on companies worldwide, causing unprecedented disruption and uncertainty. Retailers, in particular, recognized that conventional forecasting techniques based on past sales data were insufficient for estimating sales during the COVID-19 epidemic.
Due to fluctuating demand, retailers needed to focus on forecasting future store sales across longer planning horizons and toward more precise short-term planning. Additionally, they discovered a wealth of external market data, such as COVID-19 infection rates, mobility indices (Google, Apple), demographics, and macroeconomic data that might be used as drivers to explain demand patterns and increase forecast accuracy.
Stay ahead of the industry with exclusive feature stories on the top companies, expert insights and the latest news delivered straight to your inbox. Subscribe today.
Market information to enhance the accuracy and explainability of forecasts
Retail forecasting uses an increasing amount of external data and industry knowledge via publicly available data on consumer demographics, macroeconomic indicators such as Gross Domestic Product (GDP) and interest rates, social media buzz, and worldwide trade. Additionally, as demand sensing levers, leading demand indicators such as news, product evaluations, search engine data, and website glimpse views are becoming more prevalent.
By incorporating local weather, activities occurring near their locations, and traffic conditions. Forecasting approaches are evolving away from classical time series methods and toward intelligent forecasting enabled by AI/Machine Learning and cloud computing, which can consider a diverse set of external market forces and scale to retail numbers.
These next-generation systems can take leading indicator data and generate a forecasted picture that is devoid of human bias or manipulation. All the while, retailers are continually learning which leading indicator data best predicts changes for a more precise forecast, down to the smallest detail, such as the store, item, day/hour, and consumer fulfillment choice—purchase in-store, ship from store, or click-and-collect.
AI/machine learning to combine with solid feature engineering
Engineering features are vital to achieving robust results and integral to the process. Internal or external drivers, including historical sales streams, can generate features such as seasonality, causal lags, life cycle characteristics, and trends. A causal lag feature might be an event (e.g., markdown, promotion) that affects consumer purchases many days or weeks after it occurs. Machine learning is capable of iterating over many feature combinations to construct models with improved forecast accuracy at more granular levels. Machine learning combined with rigorous feature engineering produces robust forecasts that account for various demand patterns at differing levels of granularity and across a variety of consumer channels.
More in News


