Biologists regularly create phylogenetic trees to better understand the evolutionary origins of their species of interest, and often use genomes as their data source. However, as more and more incomplete genomes are published, in many cases it may not be possible to compute genome-based phylogenetic trees due to large gaps in the assembled sequences. In addition, comparison of complete genomes may not even be desirable due to the presence of horizontally acquired and homologous genes. A decision must therefore be made about which gene, or gene combinations, should be used to compute a tree. Deflated Cladistic Information based on Total Entropy (dCITE) is proposed as an easily computed metric for measuring the cladistic information in multiple sequence alignments representing a range of taxa, without the need to first compute the corresponding trees. dCITE scores can be used to rank candidate genes or decide whether input sequences provide insufficient cladistic information, making artefactual polytomies more likely. The dCITE method can be applied to protein, nucleotide or encoded phenotypic data, so can be used to select which data-type is most appropriate, given the choice. In a series of experiments the dCITE method was compared with related measures. Then, as a practical demonstration, the ideas developed in the paper were applied to a dataset representing species from the order Campylobacterales; trees based on sequence combinations, selected on the basis of their dCITE scores, were compared with a tree constructed to mimic Multi-Locus Sequence Typing (MLST) combinations of fragments. We see that the greater the dCITE score the more likely it is that the computed phylogenetic tree will be free of artefactual polytomies. Secondly, cladistic information saturates, beyond which little additional cladistic information can be obtained by adding additional sequences. Finally, sequences with high cladistic information produce more consistent trees for the same taxa.