Breast cancer is a complex disease and even at a favourable stage, such as the pT1N0 one, it is unlikely to be understood and cured by focusing only on single gene or protein alterations determined with suboptimal technologies, as the standard clinico-pathological predictors are. Improving breast cancer treatment will require a more systematic, structured and multidimensional approach able to integrate tumour biology, disease burden and host-related factors. In this scenario, multigene predictors capturing gene-expression profiling or other molecular measurements have great potential for improving breast cancer management. Nevertheless, even if the gene signatures generated so far clearly represent a step forward in the prediction of patient outcome, they are still showing some limitations that nowadays are the basis for the development of a second generation of multigene predictors. Their strength will stand in the investigation of the tumour-surrounding stroma and tumour microenvironment, in the interrogation of different molecular subtypes of breast cancer as distinct entities and in the ability to predict both early and late relapse. It is prospected that the greater accuracy of this new wave of predictors will provide substantial support to the existing decision tools and will significantly ameliorate our current ability to define breast cancer prognosis.