Frequent production requirements are linked to effective and profitable strategies for the management of queues, e.g. input buffers of machining centres and selection of rough workpieces from central storage. As a matter of fact, management policy performances must be evaluated by focusing not only on induced queue lengths or customer waiting times, but also on the future composition of the queues themselves, this being a consequence of the dispatching rule adopted. The question acquires further importance when flexible environments and buffer capacity constraints are considered. The aim of the present study is to illustrate an analytical procedure, based on the Markov chains theory, able to forecast the composition of buffers starting from any initial configuration. Attention is given to the buffer behaviour during the transient period and the steady state: results are offered for sequential and prioritized dispatching of items, together with application examples derived from common industrial situations. In the final sections, the possible extension of the procedure is considered and outlined with regard to particular management rules, thus introducing a further development of the present research.