Distribution of size of sand grains is an important factor in characterization of unconsolidated reservoirs as well as designing sand control devices. In practice, sand grains are passed through a set of known mesh sizes by mechanical vibration and for a fixed period then the weight of sediments retained on each sieve are measured and converted into the percentage of the total sediment (PTS). This procedure is applied to all core samples and the resulted PTS data are used for characterizing grain size distribution using one of the sieve analysis procedures. The core-by-core method, for example, is one of the conventional methods that PTS data from each core sample are used individually to estimate mean, sorting and other dependent parameters to grain size distribution. In this method, applying a robust statistical method to integrate all PTS data and picking out the most probable size from all cores is a challenge. A new approach is introduced in this paper as sieve-by-sieve method, whereby the grain weight distribution data are classified based on mesh sizes (as bins) and the most probable size in each class is picked out among all cores directly and without any manipulation or averaging. In this paper, the performance of both methods are compared in a homogeneous media and a heterogeneous media. In a homogeneous media, both methods provide comparable results. However, in a heterogeneous media, the core-by-core provides too many distributions which sometimes are not conclusive but the sieve-by-sieve provides the profiles of minimum and maximum weight of retained grains, which facilitates picking out the most probable size among all cores.