In this paper, we propose a formal model of decentralized search. In particular, we aim to understand how ignoring some pieces of information during technological search is rewarding for the development of radically new products. To do so, we propose an agent-based model of simulation including both product and organization structure. The NK model is used to represent a directed “technology network”, producing a rugged fitness landscape. Organizations are represented by a directed “social network”, where communication links are the counterpart of the technological interdependences. By simulating different density of the social network (from a perfectly mirroring organization to an organization made of fully autonomous agents), we find that a relative ignorance related to the effect of technological decision (on the dependent component technologies) can be rewarded by technological breakthroughs. This is particularly true when the complexity and the number of components of the products are high.