In the current economic and environmental context, improving feed efficiency has become of primary importance. In dairy cattle, the usual way to measure feed efficiency is through residual feed intake (RFI). However, this approach, in its classical form, does not take into account the evolution of the RFI components across the lactation, inducing approximations in the results. We present a new approach that incorporates the dynamic dimension of the data. Using a multi-trait random regression model, daily milk production, live weight, dry mater intake and body condition score were investigated across the lactation. Then, at each time point, the estimated variance-covariance matrix of the animal effects can be used to predict an animal effect for intake. Its difference from the actual animal effect for intake, gives an RFI estimation at each time point. This approach was tested on historical data from the Aarhus University experimental farm (1,469 lactations out of 740 cows) that were shared within the GenTORE project. The newly estimated RFI possessed all the characteristics of a traditionally calculated RFI, with a mean at zero at each time point and a phenotypic independence from its predictors. Moreover, this methodology offers new possibilities for exploring RFI changes over the lactation. For instance, the correlation between the averaged RFI over the lactation and RFI at each time point was found to be always positive and above 0.5, and maximum around mid lactation (>0.9). In addition, the model performed reasonably well in the presence of missing data. This approach allows a dynamic estimation of the traits, free from all time-related issues inherent to the traditional RFI methodology. It can be extended and used in a genetic or genomic selection context. More investigations will be conducted pooling data from different farms or with changing diets.