Almost all phylogenetic studies utilizing fossils are faced with the problem of missing data. With the development of methods allowing total evidence phylogenies where fossils are combined with extant taxa, and molecular evidence is available, this problem is compounded. Standard methods ignore missing data and often result in poorly resolved trees, and procedures such as culling of taxa or partitioning data have been proposed to improve resolution. Here, we take a different tack, one grounded in the fact that because we do not know what the behavior of missing data would be, it is worthwhile to examine the universe of possible outcomes. The MERDA value is the frequency with which a particular clade is recovered in replicated analyses where missing observations are replaced randomly with observable states. A technique to de-resolve missing data-dependent clades is also proposed. We also show that, in published data sets, there is little obvious relationship between MERDA performance and standard measures of clade support such as Bremer and Jacknife indices.