The USDA National Agricultural Statistics Service (NASS) is developing a machine-learning forecasting model for pre-season acreage estimation that includes geo-located agricultural commodity prices. This presentation describes how this price dataset is created using automation techniques in python, including looping over data files and running geo-processing tools found in the library arcpy. The data are first scraped from formatted tables of a host website. Then, they are converted into geo-referenced points and finally inverse distance weighted (IDW) surfaces, which are gridded raster-formatted large data products. This approach has ongoing applications to estimating programs covering large land areas. The automation results in a streamlined workflow that is repeatable and automatic with increased speed. It can also be applied to any case using large spatial datasets and by many machine-learning models.