Estimating the variance of survey estimates is an important but often difficult step in their evaluation and dissemination process. Precision estimates play a key role in European surveys quality reporting, as defined in European regulations, for instance the Integrated European Social Statistics frame regulation currently under negotiation. They help users in their analysis of disseminated aggregates, especially when computed on domains (industries for business statistics, regions for household statistics). Errors in surveys may have multiple sources and causes, which are for the most part difficult to evaluate quantitatively. In survey data, sampling and nonresponse errors represent a significant part of the total error, that official handbooks strongly recommend to take into account and estimate. In France, the precision estimations performed by the National Statistical Institute Insee for its surveys incorporate the following elements: sampling design, unit nonresponse usually treated by reweighting methods, influential units treatments, calibration. Variance estimation for household surveys uses analytical formulas, taking into account the real sampling design according to which the sample has been selected. These analytical formulas are implemented in taylor-made precision estimation programs, that are developped in R. Due to the complexity of sampling designs and survey data treatments, especially for household surveys, estimating variance, that is defining the precise analytical variance formula to be used and implementing it in a statistical software, is a demanding and time-consuming task, usually performed by members of the methodological staff. However, the computation of precision estimates should also be made available to data users such as subject matters experts, so that they could as easily as possible compute the variance of the variables they wish to comment and disseminate. The Gustave package has been conceived as a tool to facilitate precision estimates computation and to change their organisation. Its goal is to lessen the charge of the methodological staff and limit it to the more technical tasks for which their expertise is needed. We will here limlit our presentation on the general principles according to which Gustave is organised. The final paper will include an example of variance computation with Gustave on the French Labour Force Survey.