Visual-Near-Infra-Red (VIS/NIR) spectroscopy has led the revolution in high-throughput phenotyping methods used to determine chemical and structural elements of organic materials. In the current state of the art, spectrophotometers used for imaging techniques are either very expensive or too large to be used as a field-operable device. In this study we developed a Sparse NIR Optimization method (SNIRO) that selects a pre-determined number of wavelengths that enable quantification of analytes in a given sample using linear regression. We compared the computed complexity time and the accuracy of SNIRO to Marten's test, to forward selection test and to LASSO all applied to the determination of protein content in corn flour and meat and octane number in diesel using publicly available datasets. In addition, for the first time, we determined the glucose content in the green seaweed Ulva sp., an important feedstock for marine biorefinery. The SNIRO approach can be used as a first step in designing a spectrophotometer that can scan a small number of specific spectral regions, thus decreasing, potentially, production costs and scanner size and enabling the development of field-operable devices for content analysis of complex organic materials.