As supply chains extend geographically, the supply chain management seems crucial for improving companies’ activities and enhancing their competitiveness. This global market liberalization creates an uncertain environment for fulfilling the deliveries on time based on the complex supply chain environment consisting of different companies acting in various stages of the supply chain process. In this context, lead time is one of the most critical measures for indicating a supply chain performance, as lead time forecast accuracy is a crucial factor of mitigating the bullwhip effect. Within the framework of the ESCALATOR project, a new innovative service is being developed aiming to dynamically forecast lead time variation, through the use of historical data and dynamic forecasts of lead time delays at the previous stages of a supply chain. This paper focuses on the validation and final selection of the factors causing delays in various stages of a supply chain, as well as the most suitable machine learning models in order to develop a service for dynamic lead time predictions at different stages of a supply chain.