Congenic strains continue to be a fundamental resource for dissecting the genetic basis of complex traits. Traditionally, genetic variants (QTLs) that account for phenotypic variation in a panel of congenic strains are sought first by comparing phenotypes for each strain to the host (reference) strain, and then by examining the results to identify a common chromosome segment that provides the best match between genotype and phenotype across the panel. However, this "common-segment" method has significant limitations, including the subjective nature of the genetic model and an inability to deal formally with strain phenotypes that do not fit the model. We propose an alternative that we call "sequential" analysis and that is based on a unique principle of QTL analysis where each strain, corresponding to a single genotype, is tested individually for QTL effects rather than testing the congenic panel collectively for common effects across heterogeneous backgrounds. A minimum spanning tree, based on principles of graph theory, is used to determine the optimal sequence of strain comparisons. For two traits in two panels of congenic strains in mice, we compared results for the sequential method with the common-segment method as well as with two standard methods of QTL analysis, namely, interval mapping and multiple linear regression. The general utility of the sequential method was demonstrated with analysis of five additional traits in congenic panels from mice and rats. Sequential analysis rigorously resolved phenotypic heterogeneity among strains in the congenic panels and found QTLs that other methods failed to detect.