Modeling and Forecasting of Enterprise-level Retail Time Series Data with Implementation in SPSS

This report describes a forecasting methodology for retail chains and consumer products manufacturers, which is based on the data management tools and the time-series modeling techniques in the SPSS Statistics. The proof-of-concept of this methodology was carried out using three years price and weekly-aggregated unit sales data for the products in a specific category (Bread) that was obtained from individual grocery stores of a retail chain in a certain market geography (Kroger retail chain in the metropolitan Denver area). This methodology can incorporate additional data on exogenous demand factors which invariably improve the forecasting accuracy; these exogenous factors include for example, holiday and seasonal effects, product delivery and inventory data, promotional and marketing information, and product drag and competition effects. The implementation of this methodology with the SPSS Statistics engine called within Python programs flow provides retail chains and consumer product manufacturers with a cost-effective and scalable enterprise-level forecasting solution across the full range of their individual products and retail outlets.