Mental representations determine how individuals and firms make decisions and have implications for individual and firm performance. However, current research does not distinguish between ways of developing representations of different complexity across distinct dimensions of a decision problem. In this study, I explore the trade-offs associated with the allocation of representation search efforts across the distinct dimensions of a decision problem – that is, with the breadth of representation search strategies. To this end, I develop a NK model of dual search over policies and representations where agents can either refine their representations broadly across dimensions or deeply in one or few dimensions of a decision problem. Results obtained with this model show that the optimal representation search breadth is contingent on the complexity of the decision environment. Contrary to previous research, intermediate levels of search breadth are associated with optimal performance only for moderate levels of complexity. Higher levels of complexity demand narrow search strategies, while broad search strategies are optimal when complexity is low. A second set of results explores the relationship between the breadth of search strategies and the optimal degree of representational complexity. I find that, counterintuitively, less accurate representations can outperform more accurate ones – i.e., that the optimal degree of representational complexity does not necessarily match the true complexity of the environment. However, I show that less accurate representations can outperform more accurate ones only for broad rather than narrow representation search strategies. These findings contribute to research on learning and adaptation in complex environments and on the role of mental representation in organisational decisions.