Phylogenomic conflicts are widespread among genomic data, with most previous studies primarily focusing on nuclear datasets instead of organellar genomes. In this study, we investigate phylogenetic conflict analyses within and between plastid and mitochondrial genomes using Potentilla as a case study. We generated three plastid datasets (coding, noncoding, and all-region) and one mitochondrial dataset (coding regions) to infer phylogenies based on concatenated and multispecies coalescent (MSC) methods. Conflict analyses were then performed using PhyParts and Quartet Sampling (QS). Both plastid and mitochondrial genomes divided the Potentilla into eight highly supported clades, two of which were newly identified in this study. While most organellar loci were uninformative for the majority of nodes (bootstrap value < 70%), PhyParts and QS detected conflicting signals within the two organellar genomes. Regression analyses revealed that conflict signals mainly occurred among shorter loci, whereas longer loci tended to be more concordant with the species tree. In addition, two significant disagreements between the two organellar genomes were detected, likely attributed to hybridization and/or incomplete lineage sorting. Our results demonstrate that mitochondrial genes can fully resolve the phylogenetic relationships among eight major clades of Potentilla and are not always linked with plastome in evolutionary history. Stochastic inferences appear to be the primary source of observed conflicts among the gene trees. We recommend that the loci with short sequence length or containing limited informative sites should be used cautiously in MSC analysis, and suggest the joint application of concatenated and MSC methods for phylogenetic inference using organellar genomes.