Accurate estimations of volatility and correlation risk represent crucial inputs in terms of investment decisions. This article presents a new way to capture the portfolio dependence by introducing a new covariance estimator called the reactive covariance model. This new model is easy to implement and comes from the generalization to the multivariate framework of the reactive volatility model introduced by Valeyre et al. (2013). We examine the properties of this new covariance estimator and present its attractive features by exploring its ability to capture the dominant factors that create changes in the correlation structure of asset returns. By comparing our model with other traditional existing covariance models, we finally examine and present the advantages of using this new model in multi-asset portfolio construction.