Andesite blocks with specific mechanical properties for facing and building stones should be chosen from fresh ones. These blocks can be recognised through major elemental oxides analysis. The number of samples that can be taken and analysed is limited. Then, major elemental oxides should be estimated at unsampled locations. In this study, block compositional kriging which satisfies constraints of the compositional data and generates estimations with a minimum prediction error variances was used. The estimated blocks were classified as exploitable and unexploitable regarding major elemental oxides' thresholds of the fresh andesite obtained from petrography and mineralogy studies.