Knowledge of certain properties of a crude oil such as saturates, aromatics, resins, and asphaltenes (SARA) contents, Conradson carbon residue (CCR), ultimate analysis (CHNS), density, and molecular weight (MW) is useful for the characterization of the oil. Multivariate statistics combined with near-infrared (NIR) spectroscopy can be a powerful tool to rapidly and accurately predict these properties. Twenty-two crude oil fractions, from Alaska North Slope (ANS), the western United States (Utah, Colorado, and Wyoming), and Venezuela were used in this study. Eleven of these samples were C25+ residual fractions while the rest were C12+ residual fractions. The objective was to develop chemometric prediction models to predict the properties of unknown fractions using a single NIR spectrum. The SARA components (HPLC), molecular weights, densities, hydrogen-to-carbon (H/C) ratios, weight percent (wt %) nitrogen, weight percent sulfur (from CHNS analysis), and weight percent Conradson carbon residue (CCR) were measured. The NIR spectra for these fractions were obtained at 20 °C. Principal component analysis (PCA) and partial least-squares (PLS) techniques were used to analyze and correlate the spectra to the measured properties. Linear correlations with R 2 values greater than 0.99 were obtained for all properties studied. The uncertainty in experimental measurements for all the properties studied was comparable with the uncertainty in predictions by the models of the respective property. Furthermore, the models were tested using samples that did not belong to the calibration set. The properties predicted for these samples were within the range described by the experimental error for the respective property.