Purchasers used spreadsheets to do forecasting based on previous season’s actuals, leading to an overstocking of low-velocity SKUs and shortages of high-selling products at the same time.
Used continuous demand sensing in lieu of periodic forecasting by constantly updating forecasts based on demand signals, not planning cycles.
Developed based on Snowflake and Cortex for AI-based forecasting. Leveraged POS transactions, clickstream, promotions calendar, and weather information to produce SKU-location-week forecasts, which would automatically be written back into the merchandising planning system.
Lowered overstock by 35%, moved buyers from spreadsheet management to exception management, and increased forecast detail from category level to SKU-location level.
