The real-time measurement of soil attributes like moisture and organic carbon may eventually be useful for making real-time decisions on fertilizer application rates, herbicide application rates, or planting depth. This work features a spectrophotometer for on-the-go acquisition of NIR reflectance spectra and investigates its potential for real-time soil attribute measurement through a series of field-scale experiments. Multiple fields were mapped with the spectrophotometer after which time the spectra were compressed using principal components. Maps of the principal components were used to display spatial variability and determine sample locations. Soil samples were acquired and analyzed using reference laboratory methods and these results were used to create calibrations for the NIR spectra. The calibrations were then used to make real-time predictions of soil attributes on subsequent fields. Analysis of the accuracy and feasibility of the real-time predictions is presented and compared to post-processed results.