Online supply chain management (OSCM) is the smart way to deal with the vast amounts of data that come in from customers in a disorganized system to meet the quantities, volumes, and types of customer packages during both delivery and pick-up phases using a new design of vehicle boxes managed by IoT and to track their requests based on scheduling requests and sorting them to make a Hamiltonian route that guarantees the shortest travel distance. The OSCM framework consists of two sequential phases. 1st phase has four recruitment stages. The 1st stage discusses exploration resources (the relationship between the client and the vehicle) using IoT to receive customers' requests (Heijunka growth radius), then moves to exploration maturity to build a one-way Hamiltonian growth route direction. The 1st stage is based on tackling a Heijunka matrix fed through deep learning to classify the matrix into many conditional clusters according to customers' request forecasting and make the prediction value, which is the stop condition of cluster radius through next three stages. This study finds that XGboost outperforms Ada-boost by 14.352 % in the prediction stage. A heuristic rule based on NWBS enhances the FP-Growth algorithm over ECLAT by 7.648 % in the classification stage. Phase II is interested in reducing load and unloading activity time. This problem describes needing more than a different service at the same point (i.e., chaotic and unstable interaction leads to unstable delivery). Therefore, the online scheduling and tracking of the logistic routing using the IoT that Smart Lean Heijunka supports will enhance the SCM, increasing the visited points by 31.2 % and improving the profit by 41 %.