The authors discuss application of geostatistical methods for three-dimensional modeling of the varved clay deposit of "Gaidukovka" (Minsk region, Belarus). The study is based on data from drilling wells, using which a spatial model of clay distribution was built using the empirical Bayesian kriging (EBK 3D) method. This method allows taking spatial correlation into account, reducing the uncertainty of interpolation and selecting the parameters of variograms automatically. The analysis showed that the exponential variogram provides the best results with the minimum values of the root-mean-square error and the optimal alignment of the forecast with the actual data. Cross-checking confirmed the validity of the model and its resistance to the variability of the initial information. The constructed three-dimensional model in voxel layer and iso-surface formats enabled visualizing the distribution of clays and identifying the main zones of industrial value. It was established that about 70 per cent of the deposits are concentrated in the central part of the field at depths of 8,5–22,5 m. The northeastern areas, where the proportion of clay layers exceeds two thirds, are promising for further development. At the same time, the north-western and southeastern zones are considered unpromising due to the low content of clay fractions. The results of the study demonstrate the effectiveness of using EBK 3D to improve the accuracy of reserve assessment and mining planning. The proposed method can be adapted to other types of solid minerals, which expands its scientific and practical significance