The paper presents empirical methodology of reducing various kinds of observations in geodetic network. A special case of reducing the observation concerns cartographic mapping. For numerical illustration and comparison of methods an application of the conformal Gauss-Krüger mapping was used. Empirical methods are an alternative to the classic differential and multi- stages methods. Numerical benefits concern in particular very long geodesics, created for example by GNSS vectors. In conventional methods the numerical errors of reduction values are significantly dependent on the length of the geodesic. The proposed empirical methods do not have this unfavorable characteristics. Reduction value is determined as a difference (or especially scaled difference) of the corresponding measures of geometric elements (distances, angles), wherein these measures are approximated independently in two spaces based on the known and corresponding approximate coordinates of the network points. Since in the iterative process of the network adjustment, coordinates of the points are systematically improved, approximated reductions also converge to certain optimal values.