Planning is known to be a difficult task. One of the approaches used to reduce this difficulty is problem decomposition under divide-and-conquer methodology. This paper introduces a new decomposition technique for planning problems in STRIPS domains which is based on the idea of landmark, that is, literals that must be true in any solution plan. These landmarks are ordered and grouped in different sequential sets named intermediate goals (IG), being each IG a subgoal to solve. Our final objective is to build an independent planning sub-problem for each IG, using £¥¤§¦ as goal state and an approximation to the state that would be reached on solving sub-problem ¨© as initial state. Once obtained all the sub-problems, a multiprocessor system will be able to solve them concurrently.