An efficient data reduction procedure based on the Kendrick mass scale has been developed for on-line liquid chromatography-high resolution mass spectral (LC-HRMS) data analysis of hydrocarbon mixtures. The mass peaks are divided into 14 nominal mass series, followed by sorting homologous series according to their Kendrick mass defects (KMDs). Identification of compound type series is based on the averaged KMD of each series, thus eliminating time-consuming procedures for the determination of elemental composition of individual peaks. The use of the averaged KMD of the series also increases the accuracy of compound type identification. The efficiency of the new procedure has been demonstrated when applied to LC-HRMS analysis of heavy petroleum fractions where a large volume of data is present in each scan.