Scalable probabilistic forecasting in retail with gradient boosted trees: A practitioner’s approach Long, Xueying Bui, Quang Oktavian, Grady F. Schmidt, Daniel Bergmeir, Christoph Norbert Godahewa, Rakshitha Per Lee, Seong Zhao, Kaifeng Condylis, Paul Probabilistic forecasting Gradient boosted trees Global models The recent M5 competition has advanced the state-of-the-art in retail forecasting. However, there are important differences between the competition challenge and the challenges we face in a large e-commerce company. The datasets in our scenario are larger (hundreds of thousands of time series), and e-commerce can afford to have a larger stock assortment than brick-and-mortar retailers, leading to more intermittent data. To scale to larger dataset sizes with feasible computational effort, we investigate a two-layer hierarchy, namely the decision level with product unit sales and an aggregated level, e.g., through warehouse-product aggregation, reducing the number of series and degree of intermittency. We propose a top-down approach to forecasting at the aggregated level, and then disaggregate to obtain decision-level forecasts. Probabilistic forecasts are generated under distributional assumptions. The proposed scalable method is evaluated on both a large proprietary dataset, as well as the publicly available Corporación Favorita and M5 datasets. We are able to show the differences in characteristics of the e-commerce and brick-and-mortar retail datasets. Notably, our top-down forecasting framework enters the top 50 of the original M5 competition, even with models trained at a higher level under a much simpler setting. 2024-11-27T12:07:37Z 2024-11-27T12:07:37Z 2024-11-12 journal article Long, X. et. al. Int. J. Production Economics 279 (2025) 109449. [https://doi.org/10.1016/j.ijpe.2024.109449] https://hdl.handle.net/10481/97464 10.1016/j.ijpe.2024.109449 eng http://creativecommons.org/licenses/by/4.0/ open access Atribución 4.0 Internacional Elsevier