This paper addresses the two problems that arise in evaluating dimension-stone quarries: absence of a single quality variable and classification of blocks of stone. In constructing the single variable, an approach considering multivariate spatial relations of a number of variables that affect quality of dimension-stone is proposed. The multivariate spatial relations are defined by variogram matrices. In classification of the blocks, a decision analysis approach that rests on two key concepts, cumulative conditional distribution function and loss function, is introduced. The approaches are applied to an andesite quarry located in Ankara, Turkey: one level of the quarry is chosen to be a test area from which 108 rock samples are taken at 20 m regular intervals. The rock samples are tested for some of the mechanical properties of andesite that control the economics of extraction process, and a quality index is produced using test data. The application area is then divided into 1862 5 x 5 m square blocks, and the conditional distribution functions for these blocks are estimated by indicator kriging. Finally, the blocks are classified as exploitable and non-exploitable on the basis of minimization of economic losses. (C) 2000 Elsevier Science B.V. All rights reserved.