Temporal aggregation and shared information

What is it about?

A supplier in the retail supply chain typically replenishes its retail customer's distribution center at least once a week. On the other hand, production planning is frequently done at least monthly. As one-number forecasting becomes increasingly popular to reduce functional misalignment, demand planners frequently face the issue of selecting the appropriate time buckets at which to generate their forecast. Complicating this issue is the increased popularity of a retailer sharing its point of sale data, which contains important information that improves supplier forecast accuracy. Obvious benefits (e.g., reduced noise, less intensive computing requirements) aside, pooling granular (e.g., weekly) into larger time buckets (e.g., monthly) can have unintended consequences on forecast accuracy due to how data correlates from one period to the next. This study reveals how the noise reduction benefit of aggregating a time series may not be worth the consequences of data loss, thereby providing a more fine-tuned approach to pooling data as well as using shared demand information.

Featured Image

Why is it important?

Companies in recent years have amassed an incredible amount of data at an ever-increasing speed. Increased collaboration among suppliers and retailers further multiplied information availability. This information serves a supplier's needs along many fronts, such as production and logistics planning. Salient use of temporal aggregation may allow firms to extract greater value out of their available data. Drawing from statistical and econometric theories, this study reveals a few basic factors that demand planners should observe as they make the aggregation-information source decision with regard to forecasting.