An algorithm has been developed that employs neural network technology to retrieve soil moisture from multi-wavelength satellite observations (active/passive microwave, infrared, and visible). This represents the first step in the development of a methodology aiming to combine beneficial aspects of existing retrieval schemes. Several quality metrics have been developed to assess the performance of a retrieval product on different spatial and temporal scales. Additionally, an innovative approach to estimate the retrieval uncertainty has been proposed. An information content analysis of different satellite observations showed that active microwave observations are best suited to capture the soil moisture temporal variability, while the amplitude of the surface temperature diurnal cycle is best suited to capture the spatial variability. In a synergy analysis, it has been found that through the combination of all observations the retrieval uncertainty could be reduced by 13%. Furthermore, it was found that synergy benefits are significantly larger using a data fusion approach compared to an a posteriori combination of retrieval products, supporting the combination of different retrieval methodology aspects in a single algorithm. In a comparison with model data, it was found that the proposed methodology also shows potential to be used for the evaluation of modeled soil moisture. A comparison with in situ observations showed that the algorithm is well able to capture soil moisture spatial variabilities. It was concluded that the temporal performance can be improved through incorporation of other existing retrieval approaches.