The aim of this work is to evaluate several 3D-VAR assimilation cycles in the limited areamodel ALADIN, in comparison with the dynamical adaptation of the global model (ARPEGE) analysis, and with a focus on the precipitation forecasts. We perform a detailed evaluation for a specific, well documented, test case: Intensive observing period 14 of the Mesoscale Alpine Programme (MAP) field campaign, which is well described through various MAP data. The meteorological situation was of high interest, with triggering of convection both over the Alps and over the sea, therefore it has been chosen as the framework of our case study. There are clear benefits in favour of the 3D-VAR assimilation cycles over the first hours of forecast thanks to the observations and to the preservation of small-scale features. These improvements are further enhanced with a large-scale update step added to the 3D-VAR analysis. Innovative data are used, such as relative humidity pseudoprofiles which are processed data generated from cloud classification. They bring an important information which can redesign the frontal areas. When they are used jointly with the conventional observations in the 3D-VAR, they also lead to an improvement of the precipitation scores.