This article takes remote sensing as one of measurements. The paper overviews the general methodology from remote sensing observations to data products, and categorizes the existing methods into two types: target recognition and parameter retrieval together with their features, advantages and shortcomings. Even after 50 years of continuous research, we are still lack of consistent and scientific methodology to produce data products from remote sensing observations. In the future, in order to build up scientific and structured remote sensing methods for data products, the priority should be given to further development of multi-angle, multi-temporal, multi-spectral, and multi-source as well as both active and passive remote sensing observations, so as to develop new remote sensing indices, which have obvious ecological, geographic, agronomic meanings, to promote the normalization and standardization of remote sensing methods, and to generate synthetic products based on all available remote sensing observations instead of single remote sensing observations. Big data and cloud computing will provide support for the process from remote sensing observations to data products.