Geostationary satellite observations provide continuous monitoring of clouds over regional and global scales. Cloud properties retrieved from the satellite data are needed for evaluations of current climate and weather forecasting models. As operational satellite retrieval algorithms assume a single-layered cloud in their retrievals while multi-layer clouds occur frequently in nature, the retrievals based on a single-layer assumption can result in erroneous cloud properties that are unreliable for model evaluations. The GOES-R multi-spectral imager will provide unprecedented capability for monitoring clouds at high temporal and spatial resolutions. This study presents a novel satellite retrieval algorithm that takes advantage of the multi-spectral measurements for retrieving the properties of multi-layer clouds. The algorithm applies a dual-layer cloud radiative transfer technique that accounts for overlapped ice-over-water clouds and an iterative retrieval scheme that determines the overlying ice-cloud optical depth and effective particle size and the underlying water-cloud optical depth and effective droplet radius. Additionally, an enhanced CO2-absorbing technique accounting for the presence of underlying low cloud is developed to improve the retrieval of cirrus cloud-top height when the cirrus overlaps with low clouds. The cloud-top heights of cirrus clouds that overlay low clouds are commonly underestimated by the conventional CO2-retrieval techniques. This innovative retrieval algorithm is applied to the satellite observations obtained by the Spinning Enhanced Visible and Infrared Imager (SEVIRI) and by the Moderate Resolution Imaging Spectroradiometer (MODIS). The retrieved cloud properties are compared with the coincident active remote-sensing measurements from the ground and from space. Biases due to the single-layer assumption and improvements for multi-layer clouds in the satellite-retrieved cloud properties will be presented and discussed.