Seismic facies analysis is of great significance for the detection of residual oil in a sand-shale interbed reservoir. In this study, we propose to predict spatial distribution of sand thickness over a reservoir, based on seismic facies analysis. The target reservoir is a thin sand-shale interbed layer, and the layer thickness varies between 2 and 10 m. The thickness of sand strata within the reservoir layer appears to have a fragmentary distribution in lateral space. Thin thickness and fragmentary distribution are two factors that cause difficulty in sand thickness prediction. To tackle this problem, this study adopted a three-stage strategy. First, the reservoir over the entire study area was classified into five different lithofacies, following sedimentary microfacies analysis against the characteristics of gamma-ray logging data, and the corresponding seismic responses were meticulously depicted. Then, exploiting these seismic responses, or seismic facies, the spatial distribution of the gamma-ray values was evaluated within the thin sand-shale interbed reservoir. Finally, the spatial distribution of the sand thickness was predicted according to the spatial distribution of the gamma-ray values. The prediction was conducted independently for each seismic facies, rather than in a non-discriminatory manner. Comparing the prediction to the actual evaluation derived from well-logging data demonstrated that the thickness distribution resulting from seismic data has a high accuracy, because of the facies-based analysis.