In this study, metabolite profiling of Radix Saposhnikoviae from different geographical locations was performed using high-performance liquid chromatography-electrospray ionization time-of-flight mass spectrometry (HPLC-ESI-TOFMS) and multivariate statistical analysis technique. Principle component analysis (PCA) of the data shows that these samples could be roughly separated into three groups: Guan Fangfeng, Kou Fangfeng and Chuan Fangfeng. The potential chemical markers were discovered through the loading plot of PCA. Based on accurate mass measurements and subsequent fragment ions of TOFMS after in-source collision induced dissociation, as well as matching of empirical molecular formulae with those of published components in the in-house chemical library, 10 potential markers, such as 4'-O-glucosyl-5-O-methylvisamminol, cimifugin, prim-O-glucosylcimifugin and 3'-O-angeloylhammaudol, were tentatively identified and partially verified by the available reference standards. The results of this study indicate that it is an effective and novel approach to identify traditional Chinese medicine (TCM) from different sources, and that performing quantity determination of corresponding marker compounds could optimize the quality control of TCM.