The problem of a distribution centres network construction based on the statistical data analysis of the LTL transportation company is considered. The distribution centre network is built on the basis of the demand on terminal services. Statistical criterion for selecting the number of distribution centres in the network based on the application of the network robustness principle to the disturbances in demand for services in each terminal is suggested. Demand distortions are proposed to be carried out taking into account the forecasting of future trends in demand. The simulation study on real data is carried out. The considered task consists of a terminal network where the cargo is generated to deliver other terminals. The goal is to estimate the robust number of hubs in the network which minimizes the total flows costs and is resistant to the possible flows changes in the network. The results on the real dataset are illustrated and discussed.