Data-driven models to deal with data scarcity in warehousing system design

The design of a warehousing system involves decisions on the storage mode (i.e. technology, equipment) and the plant layout, which profoundly affects the efficiency and the costs of the operations on a long-term range. To adequately support these choices, a considerable amount of data on the inventory mix (i.e. the population of stock keeping units) is required. Nevertheless, with an increase in the number of intermediaries along the supply chains, these data are often unavailable for logistics providers, who need to optimise storage design and operations while dealing with data scarcity. This paper addresses the warehouse design problem with data scarcity. It presents a hierarchical procedure (1) for data collection and organisation, and a series models (2) to aid the warehousing system designers and managers. Great emphasis is given on the type of information (i.e. weight and volume of each SKU) necessary to feed the models. Models of increasing complexity address each decision problem (e.g. SKUs clustering, rack design, the design of buffering area). A basic support-model, fed by poor data, is always provided for each decision. More sophisticated models lead to more confident results when larger datasets are available. A case study presents the results obtained by the application of these models. The monitoring protocols about on-field monitoring, set to deal with a lack of data, are provided. The comparison of the results obtained by the application of the proposed models suggests best practices to design performing storage systems.