Sampling design plays a critical role in capturing information about spatial distribution of soils, influences directly the accuracy of samples-based soil mapping. The commonly used methods are often confronted with such problems as the need for large amounts of samples and low sampling efficiency. When the resources available are inadequate for accomplishing a large volume of sampling once for all, additional sampling has to be done many times and in batches. However, the present in-batches sampling methods mostly take into account supplementality of the batches of sampling points in geographic space, which lead to overlapping of the sampling points in attribute space, affecting efficient utilization of the resources. In view of this, clustering analysis of soil environmental co-variates was performed to explore for sampling points representative of different grades of soil spatial distribution and to formulate a sampling designing method based on representativeness grade. This proposed method was applied to the study area located in Heshan Farm, Nenjiang County, Heilongjiang Province, and digital soil mapping was done using selected sampling points of different representativeness grades for verification of the methods. Results show that only a small number of sampling points of the highest grade could obtain most of the main soil types (sub-order in Chinese Soil Taxonomy) and soil mapping was high in accuracy; and addition of sampling points lower in representative grade could raise the accuracy of soil mapping; but the accuracy did not improve much when the number of sampling points added exceeded a certain level. Therefore, compared with the number of sampling points, representativeness grade of the points was more important to accuracy of soil mapping. The representativeness grade system proposed in this paper can be used as reference for sequencing of sampling points and is contributive to formulation of high-efficiency stepwise sampling design.