Knowledge of the parentage of individuals is required to address a variety of questions concerning the evolutionary dynamics of wild populations. A major advance in parentage inference in natural populations has been the use of molecular markers and the development of statistical methods to analyse these data. Cervus, one of the most widely used parentage inference programs, uses molecular data to determine parent–offspring relationships. However, Cervus does not make use of all available information: additional phenotypic information may exist predicting parent–offspring relationships, and additional genetic information may be exploited by simultaneously considering multiple types of relationships rather than just pairwise or just parent–offspring relationships. Here we reanalyse data from a wild red deer population using two programs capable of using this additional information, MasterBayes and COLONY2, and quantify the impact of these alternative approaches by comparison with a ‘known pedigree’ estimated using a larger suite of microsatellite makers for a subset of the population. The use of phenotypic information and multiple relationships increased the number of correct assignments. We highlight the differences between programs, particularly the use of population- rather than individual-level statistical confidence in Cervus. We conclude that the use of additional information allows MasterBayes and COLONY2 to assign more correct paternities, whereas their use of individual- rather than population-level confidence generates fewer erroneous assignments. We suggest that maximal information may be gained by combining outputs from different programs. Higher accuracy and completeness of pedigree information will improve parameters estimated from pedigree information in studies of natural populations.