The selection of lots of raw materials to be used in tobacco products’ blends, especially for Premium cigars, holds a high degree of subjectivity. To overcome this, it is proposed a methodology to choose, among the available lots, only those which comply with precise quality specifications. As an integral part of the approach, this work revealed the constituents and parameters that should be analyzed in the raw materials because they are important to predict cigar strength. It was also defined NIR spectroscopic models that are good enough to analyze the important properties on products and raw materials. Two batches of 27 products (accumulating 3780 cigars) and dust samples (322 in total) were evaluated. The scores of strength were regressed on 33 variables of the cigars and the smoke, as well as on 1050 NIR reflectance measurements. The reference concentrations of the significant properties were regressed on the reflectance. Calibration and validation performances were estimated for the chemosensory and NIR spectroscopic models through partial least squares (PLS 1) and support vector regression (SVR) algorithms. The ratio nicotine/tar and the relative nicotine transfer in the smoke, together with the concentration within the products of total alkaloids as nicotine, total nitrogen, and total ash, were the significant characteristics. The prediction performance of the new chemometrics models through SVR demonstrated their usefulness for this industrial context. This work contributed to define, for the first time, a methodology for choosing the lots of raw materials and managing the optimal aging time given to processed leaves.