Heuristic approach for the solution of a transport task at a lot-based delivery of a homogeneous load in points of not-crossed clusters from several bases of service is considered. The mathematical formulation of a question is consolidated to a problem of linear programming. The algorithm of its decision consists of two stages and is based on ideas of aggregation and disaggregation of points in a cluster. At the first stage the optimization model is the problem of routing of the transportation from bases through each cluster. At the same time single run from each point is accepted equal to zero. Thanks to it the optimum ring or radial route is decided with the help of one algorithm. The problem of routing is solved by the method of dummy nodes and branches allowing to visit repeatedly tops of the transport graph. The minimum time of cargo run from base in a terminal point of unloading of a cluster on the weighed graph is used as a criterion of aggregation. It allows to consider an idle time in points of a transport network and the movement between them. At the second stage optimum distribution of weight of a load between bases and clusters in the received aggregated transport graph with the arches equal to the minimum time of cargo run is carried out. This approach has allowed to solve a transport problem taking into account features of delivery of small consignments.