While a key enabler for the modern supply chain, ERP is often mistaken as the single universal solution to meet every need. Whether the result of over-aggressive vendors or self-imposed rationalization to justify the large investment, the outcome is the same: disappointment with supply chain functionality. Invariably, the initial state of euphoria wears off with the realization that ERP provides a foundational platform, not the ultimate solution, at which point companies seek to bridge gaps in functionality by augmenting with best-of-breed solutions.
Demand planning is a classic example. Forecasting modules in ERP systems, such as SAP’s Advanced Planning Optimization (APO) or JDA, rely on time-series statistical analysis methods to create seasonal demand predictions. Fourier time-series mathematical analysis was groundbreaking when first developed in 1822, and later largely replaced by Holt-Winters exponential smoothing. However, aside from a plethora of model tuning parameters, the time-series methods employed by ERP have remained essentially unchanged in decades and are ill-suited for a modern, fast-moving supply chain. It is akin to relying on telephone books in a world of mobile phones and the Internet.
Hence, forecast accuracy for the consumer products industry remains a challenge, with average weekly forecast error (mean absolute percentage error) of more than 50 percent. Findings from the Terra Technology 2013 Forecasting Benchmark Study, encompassing $130 billion in annual sales from eleven of the largest multinational consumer products companies, reveal that one-third of items have less than the two years of history required for even the most basic seasonal time-series analysis. Furthermore, industry reliance on promotional activities to drive sales resulted in three-quarters of all items being promoted, distorting sales patterns for the one-third of items that do have two or more years of history.
